<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"
     xmlns:content="http://purl.org/rss/1.0/modules/content/"
     xmlns:dc="http://purl.org/dc/elements/1.1/"
     xmlns:dcterms="http://purl.org/dc/terms/"
     xmlns:media="http://search.yahoo.com/mrss/"
     xmlns:atom="http://www.w3.org/2005/Atom"
     xmlns:cf="https://www.futureplc.com/rss/content-flags"
>
    <channel>
                    <atom:link href="https://www.itpro.com/feeds/tag/generative-ai" rel="self" type="application/rss+xml" />
                            <title><![CDATA[ Latest from ITPro in Generative-ai ]]></title>
                <link>https://www.itpro.com/artificial-intelligence/generative-ai</link>
        <description><![CDATA[ All the latest generative-ai content from the ITPro team ]]></description>
                                    <lastBuildDate>Wed, 18 Nov 2026 09:00:00 +0000</lastBuildDate>
                            <language>en</language>
                                <item>
                                                            <title><![CDATA[ Why do AI initiatives fail – and succeed? ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Enterprise AI has moved beyond isolated pilot projects and basic conversational front-ends into a critical infrastructure challenge that directly impacts corporate architecture and operating models.</p><p>vAnd yet, even as we head into 2027, across enterprises many AI initiatives stall in the pilot phase, an indictment of IT leaders who fail to enforce architectural discipline. </p><p>Gartner <a href="https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027" target="_blank"><u>predicts</u></a> that "by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls”, more than 40% of agentic AI projects will fail. </p><p>One Gartner analyst added that "most agentic AI projects right now are early-stage experiments or proof of concepts that are mostly driven by hype and are often misapplied." </p><p>An <a href="https://executive.mit.edu/blog/developing-an-effective-ai-strategy.html" target="_blank"><u>MIT Management blog post</u></a>, meanwhile, chalked up the problems to a lack of vision, which can be laid at the feet of management. </p><p>“Many organizations dabble in pilots or isolated tools, missing opportunities to scale impact across the enterprise.”</p><p>This failure rate only tells half the story, however, and the inverse is often true. Highly <a href="https://www.itpro.com/business/business-strategy/enterprises-need-to-stop-shoehorning-ai-where-it-isnt-needed"><u>successful AI projects</u></a> can connect their wins to solid IT leadership.</p><h2 id="what-does-success-look-like">What does success look like?</h2><p>Efforts by Moderna and Jubilant Ingrevia showcase how business alignment is a hallmark of good leadership. </p><p>Moderna, the organization behind the Covid-19 vaccine, underwent a <a href="https://intuitionlabs.ai/articles/moderna-ai-adoption-case-study"><u>transformation</u></a> that “... aligned the organization around AI from the top.” The biotechnology company decided to train all employees to become data-fluent, not just their specialized workers (i.e., software engineers). </p><p>Moderna's CEO "repeatedly framed AI as essential for scaling the company’s scientific goals, not for cost cutting…" </p><p>Jubilant Ingrevia, which was <a href="https://www.mckinsey.com/capabilities/operations/how-we-help-clients/how-a-digital-operational-and-skills-transformation-took-jubilant-ingrevias-business-to-the-next-level" target="_blank"><u>struggling</u></a> to move forward, illustrates another AI-integration success story. In an effort similar to Moderna’s, the specialty chemicals company focused on training its frontline employees on IT processes and worked to foster a level of ownership over AI.</p><p>Even before beginning its AI integration efforts, management had its data scientists work onsite with plant operators for months to better understand how the business worked. </p><p>The Moderna and Jubilant Ingrevia case studies showcase how leadership decisions can directly impact the success of AI implementations. </p><p>It’s not just a technology question that boils down to choosing the "right" AI model and then seeing how it can improve things. In fact, leadership must understand core workflows and processes, then look for AI solutions that fit those circumstances.  </p><p>In short, an AI implementation failure isn't automatically due to infrastructure, model or algorithm failure. Success belongs to technology leaders who stop treating AI as a standalone novelty and start engineering it as a scalable and secure enterprise capability.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/business/business-strategy/why-do-ai-initiatives-fail-and-succeed</link>
                                                                            <description>
                            <![CDATA[ IT leaders must champion AI implementation from a business perspective, not just as a technology problem to solve ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">hcrgc5odwpjvWHZ4ztjZG4</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/Su7jfDrRkaqV3sZcdmfPHc-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 18 Nov 2026 09:00:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Business Strategy]]></category>
                                                    <category><![CDATA[Business]]></category>
                                                                                                                    <dc:creator><![CDATA[ Keith Ward ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/yayfrPRvBCXMQTMQ7V5Xw5-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/Su7jfDrRkaqV3sZcdmfPHc-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Male and female IT leaders discussing data strategy in an open plan office space while observing source code on a screen.]]></media:description>                                                            <media:text><![CDATA[Male and female IT leaders discussing data strategy in an open plan office space while observing source code on a screen.]]></media:text>
                                <media:title type="plain"><![CDATA[Male and female IT leaders discussing data strategy in an open plan office space while observing source code on a screen.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/Su7jfDrRkaqV3sZcdmfPHc-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Enterprise AI has moved beyond isolated pilot projects and basic conversational front-ends into a critical infrastructure challenge that directly impacts corporate architecture and operating models.</p><p>vAnd yet, even as we head into 2027, across enterprises many AI initiatives stall in the pilot phase, an indictment of IT leaders who fail to enforce architectural discipline. </p><p>Gartner <a href="https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027" target="_blank"><u>predicts</u></a> that "by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls”, more than 40% of agentic AI projects will fail. </p><p>One Gartner analyst added that "most agentic AI projects right now are early-stage experiments or proof of concepts that are mostly driven by hype and are often misapplied." </p><p>An <a href="https://executive.mit.edu/blog/developing-an-effective-ai-strategy.html" target="_blank"><u>MIT Management blog post</u></a>, meanwhile, chalked up the problems to a lack of vision, which can be laid at the feet of management. </p><p>“Many organizations dabble in pilots or isolated tools, missing opportunities to scale impact across the enterprise.”</p><p>This failure rate only tells half the story, however, and the inverse is often true. Highly <a href="https://www.itpro.com/business/business-strategy/enterprises-need-to-stop-shoehorning-ai-where-it-isnt-needed"><u>successful AI projects</u></a> can connect their wins to solid IT leadership.</p><h2 id="what-does-success-look-like">What does success look like?</h2><p>Efforts by Moderna and Jubilant Ingrevia showcase how business alignment is a hallmark of good leadership. </p><p>Moderna, the organization behind the Covid-19 vaccine, underwent a <a href="https://intuitionlabs.ai/articles/moderna-ai-adoption-case-study"><u>transformation</u></a> that “... aligned the organization around AI from the top.” The biotechnology company decided to train all employees to become data-fluent, not just their specialized workers (i.e., software engineers). </p><p>Moderna's CEO "repeatedly framed AI as essential for scaling the company’s scientific goals, not for cost cutting…" </p><p>Jubilant Ingrevia, which was <a href="https://www.mckinsey.com/capabilities/operations/how-we-help-clients/how-a-digital-operational-and-skills-transformation-took-jubilant-ingrevias-business-to-the-next-level" target="_blank"><u>struggling</u></a> to move forward, illustrates another AI-integration success story. In an effort similar to Moderna’s, the specialty chemicals company focused on training its frontline employees on IT processes and worked to foster a level of ownership over AI.</p><p>Even before beginning its AI integration efforts, management had its data scientists work onsite with plant operators for months to better understand how the business worked. </p><p>The Moderna and Jubilant Ingrevia case studies showcase how leadership decisions can directly impact the success of AI implementations. </p><p>It’s not just a technology question that boils down to choosing the "right" AI model and then seeing how it can improve things. In fact, leadership must understand core workflows and processes, then look for AI solutions that fit those circumstances.  </p><p>In short, an AI implementation failure isn't automatically due to infrastructure, model or algorithm failure. Success belongs to technology leaders who stop treating AI as a standalone novelty and start engineering it as a scalable and secure enterprise capability.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Yesterday’s triple AI outage should be a wake-up call for enterprises ]]></title>
                                                                                                <dc:content><![CDATA[ <p>A simultaneous outage affecting OpenAI, Anthropic and xAI should act as a wake-up call for enterprises, analysts have told <em>ITPro</em>. </p><p>All three providers reported service disruption starting at about 6.30am Pacific Time on 3 September. While the outages were resolved within two-to-three, Charlie Dai, VP principal analyst at Forrester, said the incident puts the growing reliance of enterprises on <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a> in the spotlight. </p><p>“The near-simultaneous disruptions highlight that AI is increasingly becoming operational infrastructure rather than a productivity add-on,” he told <em>ITPro</em>. </p><p>“Enterprises should treat AI availability as a resilience issue, implementing multi-model strategies, fallback workflows, and business continuity plans instead of assuming frontier AI services will always be available.”</p><p><a href="https://www.itpro.com/security/downtime-hits-hard-as-organizations-battle-hidden-costs">Downtime </a>of any kind can be highly damaging for businesses, impacting finances, consumer trust, and operational efficiency. With OpenAI and Anthropic ranking among the most popular AI providers on the market, any outage raises the stakes. </p><p>“Downtime can quickly translate into lost revenue, reduced employee productivity, delayed decisions, operational disruptions, SLA breaches, and customer dissatisfaction,” Dai said. </p><p>“As AI becomes embedded in customer service, software development, knowledge management, and business processes, even short interruptions can create cascading impacts across multiple teams and customer touchpoints.</p><h2 id="openai-anthropic-and-xai-outage-timeline">OpenAI, Anthropic, and xAI outage timeline</h2><p>OpenAI, Anthropic, and xAI first reported issues between 6.23 am and 7.45 am Pacific time yesterday. </p><p>Anthropic <a href="https://status.claude.com/" target="_blank"><u>confirmed </u></a>“elevated errors on requests” for Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 at 6.23 am PT. </p><p>OpenAI, meanwhile, <a href="https://status.openai.com/incidents/01M1KWEDH417T2CF44YYHZDFCR" target="_blank"><u>reported </u></a>“elevated errors” across <a href="https://www.itpro.com/technology/artificial-intelligence/openai-just-revealed-what-people-really-use-chatgpt-for-and-70-percent-of-queries-have-nothing-to-do-with-work">ChatGPT </a>and <a href="https://www.itpro.com/technology/artificial-intelligence/openais-codex-app-is-now-available-on-macos-and-its-free-for-some-chatgpt-users-for-a-limited-time">Codex</a>. xAI also confirmed Grok was “experiencing issues”. </p><p>Based on respective status updates, all three providers resolved the issues within two-to-three hours: OpenAI marked the case as resolved at 9.15 am, while Anthropic and xAI reported fixes at 9.16 am and 10.05 am respectively.</p><h2 id="the-mystery-source">The mystery source</h2><p>With all three providers all experiencing outages in tandem, speculation over a common source was rife. Initial suggestions that a cloud or networking provider was behind the outages fell flat, however. </p><p>Neither Amazon Web Services (AWS), Microsoft Azure, nor Google Cloud reported downtime yesterday. Cloudflare, meanwhile, told <em>ITPro </em>that services were operating as expected at the time. </p><p>“Cloudflare is not experiencing any significant service disruptions at this time,” the company said. “Our services are operating normally, and any reporting that deviates from this is incorrect.”</p><p>In a <a href="https://x.com/SpaceXAI/status/2095597264043717014?s=20" target="_blank"><u>statement via X</u></a>, however, SpaceX revealed an outage at the company’s Memphis data center yesterday morning impacted Grok services. </p><p>“We are sorry for the issues you may have experienced with Grok following an outage at our Memphis compute center this morning,” the company said. </p><p>“We’d also like to apologize to our impacted compute partners. All systems have now been restored and are functioning nominally.”</p><p>As <a href="https://www.itpro.com/software/development/anthropic-claude-code-usage-limits-increase-spacex-compute-deal"><u><em>ITPro </em></u><u>reported in May</u></a>, Anthropic signed a multi-billion dollar deal with SpaceX to rent out “<a href="https://www.anthropic.com/news/higher-limits-spacex" target="_blank"><u>all of the compute capacity</u></a>” at the company’s Colossus 1 data center. The company also relies on infrastructure provided by Amazon, Google, Broadcom, Microsoft.</p><p>Some users on X have <a href="https://x.com/mark_k/status/2095782755577610614?s=20" target="_blank"><u>pointed to a potential domino effect situation</u></a> unfolding in the wake of the SpaceX outage. With Grok and Claude down, users may have flocked to OpenAI as a backup. </p><p>A spokesperson for OpenAI told <a href="https://www.wired.com/story/nobody-is-saying-why-openai-and-anthropic-had-outages-today/" target="_blank"><u><em>Wired </em></u></a>that a “routing error” started at around 7.43 am PT. <em>ITPro </em>approached OpenAI for comment, but did not receive a response by time of publication.</p><p>Dai said that the lack of clarity from each of the providers should raise concerns for enterprises and reinforces the need for greater vendor transparency. </p><p>When multiple major providers experience overlapping failures without a clearly established common cause, enterprises cannot accurately assess systemic risk, dependency concentration, or recurrence likelihood,” he told <em>ITPro</em>. </p><p>“This reinforces the importance of vendor transparency, dependency mapping, and risk assessments that extend beyond individual AI providers to the underlying infrastructure ecosystem.”</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/security/ai-is-increasingly-becoming-operational-infrastructure-rather-than-a-productivity-add-on-yesterdays-triple-ai-outage-should-be-a-wake-up-call-for-enterprises</link>
                                                                            <description>
                            <![CDATA[ Greater vendor transparency is needed in the wake of outages at OpenAI, Anthropic, and xAI ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">VvAMFenFjr62ETLTm2pEc6</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/eyFHeH5jjDEbUmBZf8PrHJ-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 04 Sep 2026 10:07:16 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:13 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/eyFHeH5jjDEbUmBZf8PrHJ-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Smartphone screen with selection of AI apps pictured in a dedicated folder titled &#039;Artificial Intelligence&#039;, including ChatGPT, Claude, Grok, Gemini and DeepSeek app symbols.]]></media:description>                                                            <media:text><![CDATA[Smartphone screen with selection of AI apps pictured in a dedicated folder titled &#039;Artificial Intelligence&#039;, including ChatGPT, Claude, Grok, Gemini and DeepSeek app symbols.]]></media:text>
                                <media:title type="plain"><![CDATA[Smartphone screen with selection of AI apps pictured in a dedicated folder titled &#039;Artificial Intelligence&#039;, including ChatGPT, Claude, Grok, Gemini and DeepSeek app symbols.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/eyFHeH5jjDEbUmBZf8PrHJ-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>A simultaneous outage affecting OpenAI, Anthropic and xAI should act as a wake-up call for enterprises, analysts have told <em>ITPro</em>. </p><p>All three providers reported service disruption starting at about 6.30am Pacific Time on 3 September. While the outages were resolved within two-to-three, Charlie Dai, VP principal analyst at Forrester, said the incident puts the growing reliance of enterprises on <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a> in the spotlight. </p><p>“The near-simultaneous disruptions highlight that AI is increasingly becoming operational infrastructure rather than a productivity add-on,” he told <em>ITPro</em>. </p><p>“Enterprises should treat AI availability as a resilience issue, implementing multi-model strategies, fallback workflows, and business continuity plans instead of assuming frontier AI services will always be available.”</p><p><a href="https://www.itpro.com/security/downtime-hits-hard-as-organizations-battle-hidden-costs">Downtime </a>of any kind can be highly damaging for businesses, impacting finances, consumer trust, and operational efficiency. With OpenAI and Anthropic ranking among the most popular AI providers on the market, any outage raises the stakes. </p><p>“Downtime can quickly translate into lost revenue, reduced employee productivity, delayed decisions, operational disruptions, SLA breaches, and customer dissatisfaction,” Dai said. </p><p>“As AI becomes embedded in customer service, software development, knowledge management, and business processes, even short interruptions can create cascading impacts across multiple teams and customer touchpoints.</p><h2 id="openai-anthropic-and-xai-outage-timeline">OpenAI, Anthropic, and xAI outage timeline</h2><p>OpenAI, Anthropic, and xAI first reported issues between 6.23 am and 7.45 am Pacific time yesterday. </p><p>Anthropic <a href="https://status.claude.com/" target="_blank"><u>confirmed </u></a>“elevated errors on requests” for Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 at 6.23 am PT. </p><p>OpenAI, meanwhile, <a href="https://status.openai.com/incidents/01M1KWEDH417T2CF44YYHZDFCR" target="_blank"><u>reported </u></a>“elevated errors” across <a href="https://www.itpro.com/technology/artificial-intelligence/openai-just-revealed-what-people-really-use-chatgpt-for-and-70-percent-of-queries-have-nothing-to-do-with-work">ChatGPT </a>and <a href="https://www.itpro.com/technology/artificial-intelligence/openais-codex-app-is-now-available-on-macos-and-its-free-for-some-chatgpt-users-for-a-limited-time">Codex</a>. xAI also confirmed Grok was “experiencing issues”. </p><p>Based on respective status updates, all three providers resolved the issues within two-to-three hours: OpenAI marked the case as resolved at 9.15 am, while Anthropic and xAI reported fixes at 9.16 am and 10.05 am respectively.</p><h2 id="the-mystery-source">The mystery source</h2><p>With all three providers all experiencing outages in tandem, speculation over a common source was rife. Initial suggestions that a cloud or networking provider was behind the outages fell flat, however. </p><p>Neither Amazon Web Services (AWS), Microsoft Azure, nor Google Cloud reported downtime yesterday. Cloudflare, meanwhile, told <em>ITPro </em>that services were operating as expected at the time. </p><p>“Cloudflare is not experiencing any significant service disruptions at this time,” the company said. “Our services are operating normally, and any reporting that deviates from this is incorrect.”</p><p>In a <a href="https://x.com/SpaceXAI/status/2095597264043717014?s=20" target="_blank"><u>statement via X</u></a>, however, SpaceX revealed an outage at the company’s Memphis data center yesterday morning impacted Grok services. </p><p>“We are sorry for the issues you may have experienced with Grok following an outage at our Memphis compute center this morning,” the company said. </p><p>“We’d also like to apologize to our impacted compute partners. All systems have now been restored and are functioning nominally.”</p><p>As <a href="https://www.itpro.com/software/development/anthropic-claude-code-usage-limits-increase-spacex-compute-deal"><u><em>ITPro </em></u><u>reported in May</u></a>, Anthropic signed a multi-billion dollar deal with SpaceX to rent out “<a href="https://www.anthropic.com/news/higher-limits-spacex" target="_blank"><u>all of the compute capacity</u></a>” at the company’s Colossus 1 data center. The company also relies on infrastructure provided by Amazon, Google, Broadcom, Microsoft.</p><p>Some users on X have <a href="https://x.com/mark_k/status/2095782755577610614?s=20" target="_blank"><u>pointed to a potential domino effect situation</u></a> unfolding in the wake of the SpaceX outage. With Grok and Claude down, users may have flocked to OpenAI as a backup. </p><p>A spokesperson for OpenAI told <a href="https://www.wired.com/story/nobody-is-saying-why-openai-and-anthropic-had-outages-today/" target="_blank"><u><em>Wired </em></u></a>that a “routing error” started at around 7.43 am PT. <em>ITPro </em>approached OpenAI for comment, but did not receive a response by time of publication.</p><p>Dai said that the lack of clarity from each of the providers should raise concerns for enterprises and reinforces the need for greater vendor transparency. </p><p>When multiple major providers experience overlapping failures without a clearly established common cause, enterprises cannot accurately assess systemic risk, dependency concentration, or recurrence likelihood,” he told <em>ITPro</em>. </p><p>“This reinforces the importance of vendor transparency, dependency mapping, and risk assessments that extend beyond individual AI providers to the underlying infrastructure ecosystem.”</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Cheap Chinese AI models could spell trouble for US big tech ]]></title>
                                                                                                <dc:content><![CDATA[ <p>The rising quality of <a href="https://www.itpro.com/technology/artificial-intelligence/should-businesses-consider-using-chinese-ai-models">cheaper Chinese AI models</a> could wreak havoc on the US tech industry, according to Juniper Research. </p><p><a href="https://www.juniperresearch.com/resources/free-research/beyond-the-headlines-what-the-ai-bubble-really-means/" target="_blank"><u>Analysis from the consultancy</u></a> suggests leading US tech giants could find themselves losing out to Chinese rivals offering alternatives in a bid to break the industry’s global dominance. </p><p>Juniper Research pointed to figures from OpenRouter, which show that just 30% of the work on that platform was through AI models offered by leading providers such as OpenAI, Google, or Anthropic. That marks a steep fall from 70% last year. </p><p>The report noted that Chinese models typically run 60% to 90% cheaper than rivals from Anthropic and OpenAI. The pricing challenge from China was first faced last year when <a href="https://www.itpro.com/technology/artificial-intelligence/chinese-ai-firm-deepseek-has-silicon-valley-flustered"><u>DeepSeek hit the market</u></a>, tanking tech stocks briefly by offering high-level performance at a much lower cost. </p><p>This year, Alibaba's <a href="https://www.itpro.com/technology/artificial-intelligence/three-open-source-large-language-models-you-can-use-today">Qwen models</a> had overtaken Meta's Llama as the most downloaded open model system, the report noted. </p><p>February marked the "real crossover into cheaper inference", Juniper added, with Chinese models processing 4.12 trillion tokens on OpenRouter vs 2.94 trillion for American models. </p><p>That's all down to lower prices for capable-enough open models, the consultancy noted.</p><p>“Open weight models are closing the capability gap with frontier models at a fraction of the cost," said Juniper Senior Analyst Jawad Jahan.</p><p>The trend is also being aided by more efficient designs, such as Mixture of Experts (MoE) models, which only use necessary parameters for a given query. </p><p>"This cuts compute-per-query without shrinking the model’s capability," the report noted. "DeepSeek is the cleanest example of this. Its V3 models carry 671 billion parameters in total, but only 37 billion of them fire for any given token. This led to a cut of V3.2’s price by more than half, as a result of lowering the cost of long-context inference."</p><h2 id="cutting-costs-with-chinese-ai-models">Cutting costs with Chinese AI models</h2><p>According to Juniper, DeepSeek's V4 Flash cost $0.14 per million input tokens against $5.00 for <a href="https://www.itpro.com/security/openai-expands-daybreak-cyber-program-new-tools-partnerships-and-a-cyber-focused-gpt-5-5-aim-to-help-patch-the-world">OpenAI’s GPT-5.5</a>, while Claude Opus 4.8 was at the time charging about $5 for input and $25 for output per million tokens. </p><p>"This pricing gap has existed ever since the release of the first Chinese open-sourced model that was able to rival frontier organizations in the West," the report noted. </p><p>Shifting demand has altered the picture, however, with programming rising from 11% to more than half of OpenRouter's token volume. Agentic workloads also account for the majority of output, Juniper said. </p><p>"Given the large call frequency for agentic workloads, per-token pricing starts to become a real budget consideration and constraint," the report noted.</p><h2 id="funding-the-ai-roll-out">Funding the AI roll-out</h2><p>That raises a problem, one that Juniper refers to as an "existential risk": can we afford all the borrowed billions being spent on data centers via "complex financing agreements" if customers stop paying for the very best models and instead choose cheaper Chinese options? </p><p>The firm argued that if the price of running AI falls too low, companies will lose the ability to make revenue from it to pay for all this investment. </p><p>That concern isn't new, either. Investors have been <a href="https://www.itpro.com/business/business-strategy/google-clouds-record-results-cant-quiet-concerns-on-ai-spending-and-model-release-timelines"><u>questioning the massive spending on data centers</u></a>, with Juniper noting that the four major hyperscalers are expected to spend $725 billion on capital expenditure this year, up by 77% compared to the year prior.</p><p>Beyond the raw figures, Juniper pointed to the convoluted ways American companies are gathering the funds to pay for such projects, largely spurred by the AI infrastructure spending outpacing available cashflow and revenue. </p><p>The firm noted that this year the top five hyperscalers will spend more than a trillion dollars on infrastructure for AI. </p><p>"This figure exceeds their combined earnings and free cashflow," the report said. </p><p>Juniper claimed this has driven a rise in "off balance sheet financing" — in which companies don't borrow directly but build standalone legal entities to carry debt, later leasing the infrastructure back to the hyperscaler.</p><p>Circular financing is also a concern, which Juniper defines as a supplier investing in a company with the understanding the recipient will spend those funds buying the supplier's products. </p><p>The consultancy pointed to OpenAI's deals with Oracle and Nvidia as prime examples.</p><h2 id="chinese-challenge-to-funding-structures">Chinese challenge to funding structures</h2><p>If revenue from AI starts to shift to China, these spending levels and funding structures could prove problematic, the analyst firm noted. </p><p>"If this trend continues, the inference revenue underwriting the Western datacentre build-out weakens, and, correspondingly, the financing structures resting on that revenue,” Jahan commented.</p><p>That's exacerbated by the fact Chinese companies don't face the same issues, with funding either so far provided by external financial businesses, particularly in the case of DeepSeek. </p><p>Similarly, they aren’t beholden to a business model that sees cloud providers like Alibaba and Baidu effectively give the model away, then sell the compute to make use of it. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/security/open-weight-models-are-closing-the-capability-gap-with-frontier-models-at-a-fraction-of-the-cost-cheap-chinese-ai-models-could-spell-trouble-for-us-big-tech</link>
                                                                            <description>
                            <![CDATA[ Powerful new open weight models from Chinese providers could undercut AI revenues and unbalance funding models for big tech's infrastructure rollout ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">94ag4G4VKN5objAr8n2hnk</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/B2K9HhHgVDEXtjMsMYMowg-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 03 Sep 2026 15:20:15 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:42 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/B2K9HhHgVDEXtjMsMYMowg-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[The flag of the People&#039;s Republic of China (PRC) and United States pictured side-by-side.]]></media:description>                                                            <media:text><![CDATA[The flag of the People&#039;s Republic of China (PRC) and United States pictured side-by-side.]]></media:text>
                                <media:title type="plain"><![CDATA[The flag of the People&#039;s Republic of China (PRC) and United States pictured side-by-side.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/B2K9HhHgVDEXtjMsMYMowg-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>The rising quality of <a href="https://www.itpro.com/technology/artificial-intelligence/should-businesses-consider-using-chinese-ai-models">cheaper Chinese AI models</a> could wreak havoc on the US tech industry, according to Juniper Research. </p><p><a href="https://www.juniperresearch.com/resources/free-research/beyond-the-headlines-what-the-ai-bubble-really-means/" target="_blank"><u>Analysis from the consultancy</u></a> suggests leading US tech giants could find themselves losing out to Chinese rivals offering alternatives in a bid to break the industry’s global dominance. </p><p>Juniper Research pointed to figures from OpenRouter, which show that just 30% of the work on that platform was through AI models offered by leading providers such as OpenAI, Google, or Anthropic. That marks a steep fall from 70% last year. </p><p>The report noted that Chinese models typically run 60% to 90% cheaper than rivals from Anthropic and OpenAI. The pricing challenge from China was first faced last year when <a href="https://www.itpro.com/technology/artificial-intelligence/chinese-ai-firm-deepseek-has-silicon-valley-flustered"><u>DeepSeek hit the market</u></a>, tanking tech stocks briefly by offering high-level performance at a much lower cost. </p><p>This year, Alibaba's <a href="https://www.itpro.com/technology/artificial-intelligence/three-open-source-large-language-models-you-can-use-today">Qwen models</a> had overtaken Meta's Llama as the most downloaded open model system, the report noted. </p><p>February marked the "real crossover into cheaper inference", Juniper added, with Chinese models processing 4.12 trillion tokens on OpenRouter vs 2.94 trillion for American models. </p><p>That's all down to lower prices for capable-enough open models, the consultancy noted.</p><p>“Open weight models are closing the capability gap with frontier models at a fraction of the cost," said Juniper Senior Analyst Jawad Jahan.</p><p>The trend is also being aided by more efficient designs, such as Mixture of Experts (MoE) models, which only use necessary parameters for a given query. </p><p>"This cuts compute-per-query without shrinking the model’s capability," the report noted. "DeepSeek is the cleanest example of this. Its V3 models carry 671 billion parameters in total, but only 37 billion of them fire for any given token. This led to a cut of V3.2’s price by more than half, as a result of lowering the cost of long-context inference."</p><h2 id="cutting-costs-with-chinese-ai-models">Cutting costs with Chinese AI models</h2><p>According to Juniper, DeepSeek's V4 Flash cost $0.14 per million input tokens against $5.00 for <a href="https://www.itpro.com/security/openai-expands-daybreak-cyber-program-new-tools-partnerships-and-a-cyber-focused-gpt-5-5-aim-to-help-patch-the-world">OpenAI’s GPT-5.5</a>, while Claude Opus 4.8 was at the time charging about $5 for input and $25 for output per million tokens. </p><p>"This pricing gap has existed ever since the release of the first Chinese open-sourced model that was able to rival frontier organizations in the West," the report noted. </p><p>Shifting demand has altered the picture, however, with programming rising from 11% to more than half of OpenRouter's token volume. Agentic workloads also account for the majority of output, Juniper said. </p><p>"Given the large call frequency for agentic workloads, per-token pricing starts to become a real budget consideration and constraint," the report noted.</p><h2 id="funding-the-ai-roll-out">Funding the AI roll-out</h2><p>That raises a problem, one that Juniper refers to as an "existential risk": can we afford all the borrowed billions being spent on data centers via "complex financing agreements" if customers stop paying for the very best models and instead choose cheaper Chinese options? </p><p>The firm argued that if the price of running AI falls too low, companies will lose the ability to make revenue from it to pay for all this investment. </p><p>That concern isn't new, either. Investors have been <a href="https://www.itpro.com/business/business-strategy/google-clouds-record-results-cant-quiet-concerns-on-ai-spending-and-model-release-timelines"><u>questioning the massive spending on data centers</u></a>, with Juniper noting that the four major hyperscalers are expected to spend $725 billion on capital expenditure this year, up by 77% compared to the year prior.</p><p>Beyond the raw figures, Juniper pointed to the convoluted ways American companies are gathering the funds to pay for such projects, largely spurred by the AI infrastructure spending outpacing available cashflow and revenue. </p><p>The firm noted that this year the top five hyperscalers will spend more than a trillion dollars on infrastructure for AI. </p><p>"This figure exceeds their combined earnings and free cashflow," the report said. </p><p>Juniper claimed this has driven a rise in "off balance sheet financing" — in which companies don't borrow directly but build standalone legal entities to carry debt, later leasing the infrastructure back to the hyperscaler.</p><p>Circular financing is also a concern, which Juniper defines as a supplier investing in a company with the understanding the recipient will spend those funds buying the supplier's products. </p><p>The consultancy pointed to OpenAI's deals with Oracle and Nvidia as prime examples.</p><h2 id="chinese-challenge-to-funding-structures">Chinese challenge to funding structures</h2><p>If revenue from AI starts to shift to China, these spending levels and funding structures could prove problematic, the analyst firm noted. </p><p>"If this trend continues, the inference revenue underwriting the Western datacentre build-out weakens, and, correspondingly, the financing structures resting on that revenue,” Jahan commented.</p><p>That's exacerbated by the fact Chinese companies don't face the same issues, with funding either so far provided by external financial businesses, particularly in the case of DeepSeek. </p><p>Similarly, they aren’t beholden to a business model that sees cloud providers like Alibaba and Baidu effectively give the model away, then sell the compute to make use of it. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Should businesses consider using Chinese AI models? ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Chinese AI models are emerging as an increasingly capable alternative to their US-based peers. Recently, Nvidia CEO Jensen Huang <a href="https://www.itpro.com/software/open-source/these-chinese-models-are-excellent-nvidia-ceo-jensen-huang-hails-powerful-new-chinese-ai-models-like-kimi-k3-and-says-dont-be-put-off-by-security-misconceptions"><u>hailed the capabilities</u></a> of Chinese open source AI models amidst growing interest in low-cost options for enterprises. </p><p>At the same time, Hugging Face used an open-weight Chinese AI model to help mitigate the attack by the <a href="https://www.itpro.com/technology/neural-network/after-openai-hugging-face-how-do-it-leaders-need-to-change-the-way-they-think-about-ai"><u>escaped OpenAI agent</u></a>. The Cloud Security Alliance’s (CSA’s) <a href="https://cloudsecurityalliance.org/artifacts/hugging-face-ciso-post-mortem"><u>post-mortem</u></a> following the OpenAI agent escape fiasco detailed why firms should embrace open source and open-weight AI models.</p><p>“The same safety guardrails that keep frontier models from being misused for attacks can also block defenders from using those models to investigate an active one, leaving organizations without a tested open-weight fallback, at a disadvantage exactly when it matters most,” according to the CSA.</p><p>Chinese AI models are highly capable and cheaper than many US alternatives. They often perform on par with closed-source models, as <em>ITPro </em><a href="https://www.itpro.com/software/open-source/open-source-ai-performance-cost-savings-proprietary-models-linux-foundation"><u>reported in November</u></a> last year. </p><p>Yet experts are cautious about the risk they pose. Should businesses consider these models and, if so, which applications can they be used for?</p><h2 id="model-benefits">Model benefits </h2><p>Chinese models include Moonshot AI's <a href="https://forum.moonshot.ai/t/kimi-k3-is-here-our-most-capable-model/480"><u>Kimi</u></a>, Alibaba's <a href="https://qwen.ai/home"><u>Qwen</u></a>, DeepSeek, and Z.ai’s GLM. In most cases, these are best described as <a href="https://opensource.org/ai/open-weights"><u>open-weight</u></a> rather than open source, meaning they can be used by anyone and model weights and inference code are publicly available for download, but the complete training datasets and foundational code remain private. </p><p>Experts think Chinese model capabilities are impressive and improving all the time. “Their reviews and ratings show their capacity is moving toward matching the biggest and best of US frontier AI models,” says Amanda Brock, CEO at OpenUK. </p><p>At the same time, the industry is recognising the biggest <a href="https://www.itpro.com/security/why-patching-velocity-matters-as-claude-mythos-supercharges-vulnerability-discovery"><u>frontier models</u></a> are “not necessarily the best for particular tasks”, says Brock. </p><p>“The open models have reached a point where they are freely sharing some of the tech equivalent to what the closed model companies charge a subscription for and are also freely available to be iteratively developed upon.”</p><p>Capability-wise, the latest generation of Chinese AI models are “genuinely impressive”, says Assaf Morag, cybersecurity researcher at Flare. </p><p>“Based on the benchmarks and independent evaluations available today, many of these models are performing at a level comparable to other leading frontier AI models while often offering lower deployment costs and more open access.”</p><p>Among the benefits, they offer “strong reasoning, coding capabilities and large-context windows”, says Oliver Simonnet, lead cybersecurity researcher at CultureAI. He believes open-weight – and in some cases open source – models also provide better control over deployment, customisation and data residency when hosted within an organization's own infrastructure.</p><p>Chinese labs have said directly that <a href="https://hongkongfp.com/2026/08/10/how-chinese-ai-is-driving-price-competition-among-us-labs/"><u>market share</u></a> matters more than near-term revenue, releasing full weights and technical reports so developers worldwide can adopt and adapt the models freely. </p><p>At a time when enterprises are weighing up the often hefty cost of AI, Sai Molige, senior manager of threat hunting at Forescout, says models such as these now cost between 60% and 90% less to run. “That price gap, not parity on trust or security, is what's driving a real shift in where US developers send their workloads.”</p><h2 id="weighing-up-the-risks">Weighing up the risks</h2><p>Yet some critics have suggested the tools could be used as a ‘backdoor’ for Chinese intelligence services.</p><p><a href="https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi"><u><em>Axios </em></u><u>previously reported</u></a> that the White House could consider imposing restrictions or tight conditions on US firms working with these models. The US government has already <a href="https://theconversation.com/legally-or-not-the-us-government-is-controlling-global-access-to-the-worlds-most-powerful-ai-286118"><u>restricted</u></a> the use of home-grown AI in other countries. </p><p>Nvidia’s Huang thinks firms shouldn’t be put off by security “misconceptions” around Chinese open-weight and open source models. </p><p>Yet beyond the obvious scare-mongering, these models do pose some risks around data privacy and national security.</p><p>Simonnet thinks data privacy is a valid concern: Chinese data-storage laws, political censorship and bias, and training-data uncertainties “remain prominent issues”, he tells <em>ITPro.</em></p><p>The largest risks are often around data governance, supply-chain trust, compliance obligations, and operational security – and this is not necessarily the model weights themselves, says Morag. </p><p>“From a cybersecurity perspective, every external AI service introduces another third-party dependency. The same due diligence applied to cloud providers or SaaS platforms should also apply to AI models, regardless of whether they originate in China, the US or Europe.”</p><p>Self-hosting the models can reduce some of these privacy risks. However, this doesn't remove the risk of model biases or technical vulnerabilities. Indeed, it requires organizations to further secure and maintain the model themselves, which “adds an extra layer of security challenges”, according to Simonnet.</p><p>However, when assessing the risks, the discussion should move beyond simply asking whether a model is Chinese, says Morag. “Organizations need to evaluate where inference occurs, what data leaves their environment, who operates the infrastructure, how updates are delivered, and whether the model can be independently audited.”</p><h2 id="the-verdict">The verdict </h2><p>Despite posing some risks, experts say Chinese models are an option in many cases when compared to frontier alternatives. Open-weight models offer organizations “substantially more control”, says Morag. </p><p>“They can be deployed inside private infrastructure, reducing the need to move sensitive corporate information outside the boundaries of the organization to third-party providers. They also enable independent security testing and auditing, which is difficult or impossible with closed commercial APIs.”</p><p>Closed models, however, generally provide stronger vendor support, as well as managed security controls and predictable service levels, according to Morag. “For many enterprises, the decision should be based on governance requirements and operational maturity, rather than geography alone.”</p><p>Overall, Simonnet has a positive view of Chinese models. He points out they can deliver “strong performance with fewer restrictions at a smaller price”, citing the example of the Hugging Face incident. </p><p>Open-weight models do provide greater control and a way to keep sensitive data within an organization when they are privately hosted, says Simonnet. However, self-hosting requires additional technical expertise and security oversight, he concedes.</p><p>At the same time, he warns that confidential, regulated, or other sensitive data “should not be entered into these services without the proper security, privacy, and governance controls being in place”. </p><p>But for general-purpose, non-sensitive work at scale, the cost case is strong, provided the deployment is self-hosted and audited, says Molige. </p><p>“For code headed for production or government-adjacent systems, companies should scan generated code regardless of which model wrote it, rather than trusting output by source.”</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/should-businesses-consider-using-chinese-ai-models</link>
                                                                            <description>
                            <![CDATA[ Chinese AI models are highly capable and often low-cost, but experts are cautious about the risks they pose. Should businesses consider these models, and if so, which applications can they be used for? ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">Gw5cak4erURwot3tzYjoZM</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/bLSsYgswXfVWXGiXgSJvj8-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 03 Sep 2026 07:00:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:03 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Kate O&#039;Flaherty ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/LUULv6n7VJ3BHPnaoLHHdg-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/bLSsYgswXfVWXGiXgSJvj8-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Conceptual image of a large-scale futuristic data center on a peach grid showcasing a prominent glowing AI cube cluster emerging from an orange container, with extensive colorful wiring linking to surrounding illuminated server racks in precise formations, representing advanced artificial intelligence infrastructure, machine learning networks, and interconnected digital systems in a high-tech conceptual 3D scene with vibrant effects.]]></media:description>                                                            <media:text><![CDATA[Conceptual image of a large-scale futuristic data center on a peach grid showcasing a prominent glowing AI cube cluster emerging from an orange container, with extensive colorful wiring linking to surrounding illuminated server racks in precise formations, representing advanced artificial intelligence infrastructure, machine learning networks, and interconnected digital systems in a high-tech conceptual 3D scene with vibrant effects.]]></media:text>
                                <media:title type="plain"><![CDATA[Conceptual image of a large-scale futuristic data center on a peach grid showcasing a prominent glowing AI cube cluster emerging from an orange container, with extensive colorful wiring linking to surrounding illuminated server racks in precise formations, representing advanced artificial intelligence infrastructure, machine learning networks, and interconnected digital systems in a high-tech conceptual 3D scene with vibrant effects.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/bLSsYgswXfVWXGiXgSJvj8-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Chinese AI models are emerging as an increasingly capable alternative to their US-based peers. Recently, Nvidia CEO Jensen Huang <a href="https://www.itpro.com/software/open-source/these-chinese-models-are-excellent-nvidia-ceo-jensen-huang-hails-powerful-new-chinese-ai-models-like-kimi-k3-and-says-dont-be-put-off-by-security-misconceptions"><u>hailed the capabilities</u></a> of Chinese open source AI models amidst growing interest in low-cost options for enterprises. </p><p>At the same time, Hugging Face used an open-weight Chinese AI model to help mitigate the attack by the <a href="https://www.itpro.com/technology/neural-network/after-openai-hugging-face-how-do-it-leaders-need-to-change-the-way-they-think-about-ai"><u>escaped OpenAI agent</u></a>. The Cloud Security Alliance’s (CSA’s) <a href="https://cloudsecurityalliance.org/artifacts/hugging-face-ciso-post-mortem"><u>post-mortem</u></a> following the OpenAI agent escape fiasco detailed why firms should embrace open source and open-weight AI models.</p><p>“The same safety guardrails that keep frontier models from being misused for attacks can also block defenders from using those models to investigate an active one, leaving organizations without a tested open-weight fallback, at a disadvantage exactly when it matters most,” according to the CSA.</p><p>Chinese AI models are highly capable and cheaper than many US alternatives. They often perform on par with closed-source models, as <em>ITPro </em><a href="https://www.itpro.com/software/open-source/open-source-ai-performance-cost-savings-proprietary-models-linux-foundation"><u>reported in November</u></a> last year. </p><p>Yet experts are cautious about the risk they pose. Should businesses consider these models and, if so, which applications can they be used for?</p><h2 id="model-benefits">Model benefits </h2><p>Chinese models include Moonshot AI's <a href="https://forum.moonshot.ai/t/kimi-k3-is-here-our-most-capable-model/480"><u>Kimi</u></a>, Alibaba's <a href="https://qwen.ai/home"><u>Qwen</u></a>, DeepSeek, and Z.ai’s GLM. In most cases, these are best described as <a href="https://opensource.org/ai/open-weights"><u>open-weight</u></a> rather than open source, meaning they can be used by anyone and model weights and inference code are publicly available for download, but the complete training datasets and foundational code remain private. </p><p>Experts think Chinese model capabilities are impressive and improving all the time. “Their reviews and ratings show their capacity is moving toward matching the biggest and best of US frontier AI models,” says Amanda Brock, CEO at OpenUK. </p><p>At the same time, the industry is recognising the biggest <a href="https://www.itpro.com/security/why-patching-velocity-matters-as-claude-mythos-supercharges-vulnerability-discovery"><u>frontier models</u></a> are “not necessarily the best for particular tasks”, says Brock. </p><p>“The open models have reached a point where they are freely sharing some of the tech equivalent to what the closed model companies charge a subscription for and are also freely available to be iteratively developed upon.”</p><p>Capability-wise, the latest generation of Chinese AI models are “genuinely impressive”, says Assaf Morag, cybersecurity researcher at Flare. </p><p>“Based on the benchmarks and independent evaluations available today, many of these models are performing at a level comparable to other leading frontier AI models while often offering lower deployment costs and more open access.”</p><p>Among the benefits, they offer “strong reasoning, coding capabilities and large-context windows”, says Oliver Simonnet, lead cybersecurity researcher at CultureAI. He believes open-weight – and in some cases open source – models also provide better control over deployment, customisation and data residency when hosted within an organization's own infrastructure.</p><p>Chinese labs have said directly that <a href="https://hongkongfp.com/2026/08/10/how-chinese-ai-is-driving-price-competition-among-us-labs/"><u>market share</u></a> matters more than near-term revenue, releasing full weights and technical reports so developers worldwide can adopt and adapt the models freely. </p><p>At a time when enterprises are weighing up the often hefty cost of AI, Sai Molige, senior manager of threat hunting at Forescout, says models such as these now cost between 60% and 90% less to run. “That price gap, not parity on trust or security, is what's driving a real shift in where US developers send their workloads.”</p><h2 id="weighing-up-the-risks">Weighing up the risks</h2><p>Yet some critics have suggested the tools could be used as a ‘backdoor’ for Chinese intelligence services.</p><p><a href="https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi"><u><em>Axios </em></u><u>previously reported</u></a> that the White House could consider imposing restrictions or tight conditions on US firms working with these models. The US government has already <a href="https://theconversation.com/legally-or-not-the-us-government-is-controlling-global-access-to-the-worlds-most-powerful-ai-286118"><u>restricted</u></a> the use of home-grown AI in other countries. </p><p>Nvidia’s Huang thinks firms shouldn’t be put off by security “misconceptions” around Chinese open-weight and open source models. </p><p>Yet beyond the obvious scare-mongering, these models do pose some risks around data privacy and national security.</p><p>Simonnet thinks data privacy is a valid concern: Chinese data-storage laws, political censorship and bias, and training-data uncertainties “remain prominent issues”, he tells <em>ITPro.</em></p><p>The largest risks are often around data governance, supply-chain trust, compliance obligations, and operational security – and this is not necessarily the model weights themselves, says Morag. </p><p>“From a cybersecurity perspective, every external AI service introduces another third-party dependency. The same due diligence applied to cloud providers or SaaS platforms should also apply to AI models, regardless of whether they originate in China, the US or Europe.”</p><p>Self-hosting the models can reduce some of these privacy risks. However, this doesn't remove the risk of model biases or technical vulnerabilities. Indeed, it requires organizations to further secure and maintain the model themselves, which “adds an extra layer of security challenges”, according to Simonnet.</p><p>However, when assessing the risks, the discussion should move beyond simply asking whether a model is Chinese, says Morag. “Organizations need to evaluate where inference occurs, what data leaves their environment, who operates the infrastructure, how updates are delivered, and whether the model can be independently audited.”</p><h2 id="the-verdict">The verdict </h2><p>Despite posing some risks, experts say Chinese models are an option in many cases when compared to frontier alternatives. Open-weight models offer organizations “substantially more control”, says Morag. </p><p>“They can be deployed inside private infrastructure, reducing the need to move sensitive corporate information outside the boundaries of the organization to third-party providers. They also enable independent security testing and auditing, which is difficult or impossible with closed commercial APIs.”</p><p>Closed models, however, generally provide stronger vendor support, as well as managed security controls and predictable service levels, according to Morag. “For many enterprises, the decision should be based on governance requirements and operational maturity, rather than geography alone.”</p><p>Overall, Simonnet has a positive view of Chinese models. He points out they can deliver “strong performance with fewer restrictions at a smaller price”, citing the example of the Hugging Face incident. </p><p>Open-weight models do provide greater control and a way to keep sensitive data within an organization when they are privately hosted, says Simonnet. However, self-hosting requires additional technical expertise and security oversight, he concedes.</p><p>At the same time, he warns that confidential, regulated, or other sensitive data “should not be entered into these services without the proper security, privacy, and governance controls being in place”. </p><p>But for general-purpose, non-sensitive work at scale, the cost case is strong, provided the deployment is self-hosted and audited, says Molige. </p><p>“For code headed for production or government-adjacent systems, companies should scan generated code regardless of which model wrote it, rather than trusting output by source.”</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Anthropic says Claude Fable 5.1 ‘sets a new standard for coding, knowledge work, and long-running problem-solving tasks’ ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Anthropic has cut the ribbon on its Claude Fable 5.1 AI model, bringing performance enhancements and security improvements – along with new cost efficiency boosts. </p><p>According to the firm, Fable 5.1 “sets a new standard for coding, knowledge work, and long-running problem-solving tasks”. </p><p>Internal benchmark tests showed the new model had significantly higher performance capabilities than <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-just-launched-claude-fable-5-its-first-mythos-class-ai-model-but-it-has-new-safeguards-to-prevent-misuse-and-will-fall-back-to-opus-4-8-for-high-risk-queries"><u>Fable 5</u></a> across four key metrics: agentic scientific research, agentic terminal coding, multidisciplinary reasoning, and agentic coding. </p><p>Anthropic also claimed that Fable 5.1 is roughly 25% cheaper than its predecessor for “typical workloads”, providing usage is billed on a per-token basis. Anthropic said this is due to price reductions in cache reads, which involve the model reading inputs that have already been processed and stored. </p><p>“For highly agentic work, the savings will often be much larger—up to approximately 45%,” the company said in a <a href="https://www.anthropic.com/claude-fable-and-mythos-5-1" target="_blank"><u>blog post</u></a>. </p><p>These improvements come amidst rising concerns about AI-related costs, which are largely down to how the technology is now being used. The shift to agentic AI is a key factor here, mainly as agents typically consume more tokens compared to traditional chatbot-based interactions. </p><p><a href="https://signal65.com/wp-content/uploads/2026/05/Signal65-Insights_The-Economics-of-Agentic-AI.pdf" target="_blank"><u>Research from Signal65</u></a> found that agentic AI workloads consume anywhere up to fifteen-times more tokens compared to chatbots. </p><h2 id="new-safeguards-for-claude-fable-5-1">New safeguards for Claude Fable 5.1</h2><p>Anthropic also revealed a series of new safeguards for Fable 5.1. This includes new Enterprise Frontier Safeguards (EFS), a system that gives customers “complete privacy” through a zero data retention policy. </p><p>EFS works by storing data in cloud infrastructure “controlled entirely by the customer”, the company said and is expected to roll out later this year for enterprise customers. </p><p>New <a href="https://www.itpro.com/security/28133/what-is-cyber-security">cybersecurity</a>-related safeguards have also been introduced, with particular focus on reducing false positives. Anthropic said changes rolled out with Fable 5.1 deliver a 60% reduction in false positives than before. </p><p>“In part, this is because Fable 5.1 can now be used to discover software vulnerabilities – though not to develop exploits for them,” the company explained.</p><p>Prior to Fable 5.1’s release, Anthropic said it performed “extensive stress-testing” of safeguards to prevent potential jailbreaks and misuse. </p><p>“In addition to our own dynamic evaluations of their robustness, we commissioned external testing from two organisations, along with automated testing by Gray Swan,” the company said. </p><p>“As with Fable 5 and Opus 5, we have not found evidence of a critical-severity jailbreak for these safeguards.”</p><p>Fable 5.1 is available now for customers through selected cloud platforms, or through the Anthropic API. </p><p>Mythos 5.1, meanwhile, will only be available to registered partners conducting cybersecurity of life science research. This continues the company’s strategy of locking down access to selected partners due to <a href="https://www.itpro.com/security/anthropic-resumes-model-testing-after-recent-cyber-incidents-but-its-introduced-new-rules-to-improve-security">security-related concerns</a>. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/anthropic-says-claude-fable-5-1-sets-a-new-standard-for-coding-knowledge-work-and-long-running-problem-solving-tasks</link>
                                                                            <description>
                            <![CDATA[ The launch of Claude Fable 5.1 includes new security and privacy safeguards, and at a cheaper rate ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">zAkpY5wg7JZRD2zz4YdXg5</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/YLuvjPah4So7sw69Ms6dLG-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 02 Sep 2026 14:40:01 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:48 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/YLuvjPah4So7sw69Ms6dLG-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Logo of AI company Anthropic pictured on a smartphone screen placed on top of reflective surface, with company logo reflected.]]></media:description>                                                            <media:text><![CDATA[Logo of AI company Anthropic pictured on a smartphone screen placed on top of reflective surface, with company logo reflected.]]></media:text>
                                <media:title type="plain"><![CDATA[Logo of AI company Anthropic pictured on a smartphone screen placed on top of reflective surface, with company logo reflected.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/YLuvjPah4So7sw69Ms6dLG-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Anthropic has cut the ribbon on its Claude Fable 5.1 AI model, bringing performance enhancements and security improvements – along with new cost efficiency boosts. </p><p>According to the firm, Fable 5.1 “sets a new standard for coding, knowledge work, and long-running problem-solving tasks”. </p><p>Internal benchmark tests showed the new model had significantly higher performance capabilities than <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-just-launched-claude-fable-5-its-first-mythos-class-ai-model-but-it-has-new-safeguards-to-prevent-misuse-and-will-fall-back-to-opus-4-8-for-high-risk-queries"><u>Fable 5</u></a> across four key metrics: agentic scientific research, agentic terminal coding, multidisciplinary reasoning, and agentic coding. </p><p>Anthropic also claimed that Fable 5.1 is roughly 25% cheaper than its predecessor for “typical workloads”, providing usage is billed on a per-token basis. Anthropic said this is due to price reductions in cache reads, which involve the model reading inputs that have already been processed and stored. </p><p>“For highly agentic work, the savings will often be much larger—up to approximately 45%,” the company said in a <a href="https://www.anthropic.com/claude-fable-and-mythos-5-1" target="_blank"><u>blog post</u></a>. </p><p>These improvements come amidst rising concerns about AI-related costs, which are largely down to how the technology is now being used. The shift to agentic AI is a key factor here, mainly as agents typically consume more tokens compared to traditional chatbot-based interactions. </p><p><a href="https://signal65.com/wp-content/uploads/2026/05/Signal65-Insights_The-Economics-of-Agentic-AI.pdf" target="_blank"><u>Research from Signal65</u></a> found that agentic AI workloads consume anywhere up to fifteen-times more tokens compared to chatbots. </p><h2 id="new-safeguards-for-claude-fable-5-1">New safeguards for Claude Fable 5.1</h2><p>Anthropic also revealed a series of new safeguards for Fable 5.1. This includes new Enterprise Frontier Safeguards (EFS), a system that gives customers “complete privacy” through a zero data retention policy. </p><p>EFS works by storing data in cloud infrastructure “controlled entirely by the customer”, the company said and is expected to roll out later this year for enterprise customers. </p><p>New <a href="https://www.itpro.com/security/28133/what-is-cyber-security">cybersecurity</a>-related safeguards have also been introduced, with particular focus on reducing false positives. Anthropic said changes rolled out with Fable 5.1 deliver a 60% reduction in false positives than before. </p><p>“In part, this is because Fable 5.1 can now be used to discover software vulnerabilities – though not to develop exploits for them,” the company explained.</p><p>Prior to Fable 5.1’s release, Anthropic said it performed “extensive stress-testing” of safeguards to prevent potential jailbreaks and misuse. </p><p>“In addition to our own dynamic evaluations of their robustness, we commissioned external testing from two organisations, along with automated testing by Gray Swan,” the company said. </p><p>“As with Fable 5 and Opus 5, we have not found evidence of a critical-severity jailbreak for these safeguards.”</p><p>Fable 5.1 is available now for customers through selected cloud platforms, or through the Anthropic API. </p><p>Mythos 5.1, meanwhile, will only be available to registered partners conducting cybersecurity of life science research. This continues the company’s strategy of locking down access to selected partners due to <a href="https://www.itpro.com/security/anthropic-resumes-model-testing-after-recent-cyber-incidents-but-its-introduced-new-rules-to-improve-security">security-related concerns</a>. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Snowflake wants to help you cut AI costs by choosing the right model for the right task ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Snowflake has announced new dynamic model routing capabilities for flagship products along with expanded access to open-weight models for customers. </p><p>The new capabilities build on Cortex AI Gateway, according to the firm. It was unveiled in July this year and provides customers with a “unified foundation” for agent governance and routing requests. </p><p>The aim here, the company noted, is primarily focused on optimizing AI token consumption, which has become a recurring pain point for enterprises over the last 12 months. </p><p>Snowflake said the new capabilities mean Cortex AI Gateway will “automatically select" a specific model based on the quality and cost for particular tasks. </p><p>Put simply, this will see Snowflake assign more efficient models for lower-complexity or repetitive tasks. Meanwhile, tasks that require “deeper reasoning” are routed to frontier models. </p><p>This means customers can reduce unnecessary inference spend and remove the need for laborious manual model selection for each individual task, according to the firm. </p><p>Snowflake CEO Sridhar Ramaswamy said the update comes in direct response to growing concerns over <a href="https://www.itpro.com/technology/artificial-intelligence/it-leaders-are-being-stung-by-unexpected-ai-costs">AI-related costs</a>. <a href="https://www.itpro.com/technology/artificial-intelligence/could-rising-token-costs-boost-interest-in-on-premises-hardware">Token costs</a> have surged so far across 2026, while a host of major providers have switched to new <a href="https://www.itpro.com/software/development/surging-ai-costs-could-exceed-developer-salaries-by-2028-analysts-say-context-engineering-could-be-the-key-to-optimizing-token-consumption">consumption-based models</a>, resulting in increased costs for enterprises. </p><p>“Enterprises are becoming much more rigorous about the economics of AI. The question is no longer how much AI they are using, but whether that AI is translating into meaningful business value,” said Ramaswamy. </p><p>“Achieving intelligence efficiency requires the flexibility to use the best model for each task as the landscape evolves. Snowflake’s role is to absorb that complexity so customers can focus on outcomes while we optimize model choice underneath.”</p><p>Dynamic routing will also be integrated across Snowflake CoCo and Snowflake CoWork, and will be made available for third-party AI agents. </p><h2 id="snowflake-expands-model-library-with-open-weight-options">Snowflake expands model library with open-weight options </h2><p>In addition to model routing capabilities, Snowflake revealed it will expand access to several <a href="https://www.itpro.com/technology/artificial-intelligence/big-tech-faces-an-adapt-or-die-predicament-with-open-weight-ai-models">leading open models</a>, such as DeepSeek-V4-Flash 0731 and GLM-5.3. </p><p>The addition of these models adds to what is already an extensive library available for customers, Snowflake said. Enterprises can already choose from a variety of models from Anthropic, Google, Meta, Mistral, and OpenAI. </p><p>“This portfolio gives customers more freedom to choose the right combination of performance and cost for each workload,” the company said in a blog post. </p><p>Once again, the focus here is firmly on helping to manage costs by providing a broader range of options. </p><p>Internal testing conducted by the firm found that enterprises using a combination of open and proprietary models can “deliver comparable quality while materially improving efficiency”. </p><p>In one evaluation detailed by the company, agents using dynamic model routing through Cortex AI Gateway were able to build a data build tool (dbt) pipeline with up to three times greater token efficiency compared to a frontier model-only approach – all while maintaining the same quality. </p><p>“In a separate test, engineering teams completed the same number of pull requests with 25 percent greater token efficiency,” the company said in a blog post. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/enterprises-are-becoming-much-more-rigorous-about-the-economics-of-ai-snowflake-wants-to-help-you-cut-ai-costs-by-choosing-the-right-model-for-the-right-task</link>
                                                                            <description>
                            <![CDATA[ New dynamic model routing capabilities aim to help customers box clever when it comes to their AI model choice ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">4RcSMAHWj96Vy6L7UdoQL8</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/KbHt2WSvhdcPMQmCqZKL8T-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 18 Aug 2026 16:34:19 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:19 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/KbHt2WSvhdcPMQmCqZKL8T-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Snowflake CEO Sridhar Ramaswamy speaks during the Snowflake Summit 26 at Moscone Center on June 1, 2026 in San Francisco, California. ]]></media:description>                                                            <media:text><![CDATA[Snowflake CEO Sridhar Ramaswamy speaks during the Snowflake Summit 26 at Moscone Center on June 1, 2026 in San Francisco, California. ]]></media:text>
                                <media:title type="plain"><![CDATA[Snowflake CEO Sridhar Ramaswamy speaks during the Snowflake Summit 26 at Moscone Center on June 1, 2026 in San Francisco, California. ]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/KbHt2WSvhdcPMQmCqZKL8T-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Snowflake has announced new dynamic model routing capabilities for flagship products along with expanded access to open-weight models for customers. </p><p>The new capabilities build on Cortex AI Gateway, according to the firm. It was unveiled in July this year and provides customers with a “unified foundation” for agent governance and routing requests. </p><p>The aim here, the company noted, is primarily focused on optimizing AI token consumption, which has become a recurring pain point for enterprises over the last 12 months. </p><p>Snowflake said the new capabilities mean Cortex AI Gateway will “automatically select" a specific model based on the quality and cost for particular tasks. </p><p>Put simply, this will see Snowflake assign more efficient models for lower-complexity or repetitive tasks. Meanwhile, tasks that require “deeper reasoning” are routed to frontier models. </p><p>This means customers can reduce unnecessary inference spend and remove the need for laborious manual model selection for each individual task, according to the firm. </p><p>Snowflake CEO Sridhar Ramaswamy said the update comes in direct response to growing concerns over <a href="https://www.itpro.com/technology/artificial-intelligence/it-leaders-are-being-stung-by-unexpected-ai-costs">AI-related costs</a>. <a href="https://www.itpro.com/technology/artificial-intelligence/could-rising-token-costs-boost-interest-in-on-premises-hardware">Token costs</a> have surged so far across 2026, while a host of major providers have switched to new <a href="https://www.itpro.com/software/development/surging-ai-costs-could-exceed-developer-salaries-by-2028-analysts-say-context-engineering-could-be-the-key-to-optimizing-token-consumption">consumption-based models</a>, resulting in increased costs for enterprises. </p><p>“Enterprises are becoming much more rigorous about the economics of AI. The question is no longer how much AI they are using, but whether that AI is translating into meaningful business value,” said Ramaswamy. </p><p>“Achieving intelligence efficiency requires the flexibility to use the best model for each task as the landscape evolves. Snowflake’s role is to absorb that complexity so customers can focus on outcomes while we optimize model choice underneath.”</p><p>Dynamic routing will also be integrated across Snowflake CoCo and Snowflake CoWork, and will be made available for third-party AI agents. </p><h2 id="snowflake-expands-model-library-with-open-weight-options">Snowflake expands model library with open-weight options </h2><p>In addition to model routing capabilities, Snowflake revealed it will expand access to several <a href="https://www.itpro.com/technology/artificial-intelligence/big-tech-faces-an-adapt-or-die-predicament-with-open-weight-ai-models">leading open models</a>, such as DeepSeek-V4-Flash 0731 and GLM-5.3. </p><p>The addition of these models adds to what is already an extensive library available for customers, Snowflake said. Enterprises can already choose from a variety of models from Anthropic, Google, Meta, Mistral, and OpenAI. </p><p>“This portfolio gives customers more freedom to choose the right combination of performance and cost for each workload,” the company said in a blog post. </p><p>Once again, the focus here is firmly on helping to manage costs by providing a broader range of options. </p><p>Internal testing conducted by the firm found that enterprises using a combination of open and proprietary models can “deliver comparable quality while materially improving efficiency”. </p><p>In one evaluation detailed by the company, agents using dynamic model routing through Cortex AI Gateway were able to build a data build tool (dbt) pipeline with up to three times greater token efficiency compared to a frontier model-only approach – all while maintaining the same quality. </p><p>“In a separate test, engineering teams completed the same number of pull requests with 25 percent greater token efficiency,” the company said in a blog post. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Why Google just spent $10 million on Spirit Airlines data ]]></title>
                                                                                                <dc:content><![CDATA[ <p><a href="https://www.bbc.co.uk/news/articles/cqxlnrqjvzyo" target="_blank">Spirit Airlines collapsed</a> in May this year, but its data is set for a new lease of life after Google acquired it in an auction. </p><p>The tech giant recently outbid AI company Mercor to acquire the de-identified data, according to <a href="https://document.epiq11.com/document/getdocumentbycode?docId=4606206&amp;projectCode=SPJ&amp;source=DM&amp;utm_s" target="_blank">court documents</a>. Google bid $10 million, versus Mercor's $7.5 million, though reports note the bidding started around the $5 million mark. </p><p>A third bidder also took part but was not named. </p><p>A spokesperson for Google told <em>ITPro </em>the aim behind acquiring the defunct airline’s data lies in its potential use for AI training and product refinement. </p><p>"We acquired part of an enterprise dataset from Spirit Airlines, which can be helpful in improving our products and AI models,” the spokesperson said. </p><p>That same sentiment was highlighted by a spokesperson for Mercor, per reports from <a href="https://www.businessinsider.com/google-buys-spirit-airlines-data-ai-model-development-2026-8" target="_blank"><em>Business Insider</em></a>. </p><p>"Companies are sitting on decades of records that show how real work gets done, and that data is now some of the most valuable material for training and evaluating AI," the spokesperson said. </p><p>"We partner with leading companies to license their operational data to the labs building the next generation of models. Spirit was that same process applied to a bankruptcy estate." </p><h2 id="what-google-bought">What Google bought </h2><p>The data bought by Google includes all productivity and collaboration data, which the filing notes includes everything from emails to chats, spreadsheets to calendars, as well as "core business systems and business applications data". </p><p>This includes data on employee behavior and productivity, aircraft operations, and inventory, among other topics. Beyond that, Google has purchased workflow and process data from departments such as HR and marketing. </p><p>In total, Google is picking up 100 million emails, 500 million Microsoft Teams chats, 30 million lines of code – but not records of loyalty program members or other customer data. </p><p>Indeed, the data set does not include any consumer data or personal data, the filing noted, or any other information that could fall under data protection laws. The buyer was required to agree to not attempt to re-identify the data and link it to any person. </p><p>The buyer is allowed to transfer the data set to third parties, however. </p><p>"We will not receive any personal information from this dataset," the Google spokesperson added. "Any data we receive will be rigorously scrubbed of any personally identifiable information by a third party before receipt."</p><p>Alongside the data, Google's $10 million also buys it all of Spirit's internally developed software and applications, including all the <a href="https://www.itpro.com/software/open-source/what-red-hats-source-code-restrictions-mean-for-businesses">source code</a>, plugins, data files, libraries, <a href="https://www.itpro.com/tag/application-programming-interface">APIs</a> and documentation. </p><h2 id="the-battle-for-data">The battle for data</h2><p>The battle for the business data reveals how desperate AI developers are for fresh material to feed AI, in particular for training models, though Google didn't specifically say that was the purpose of the data set. </p><p>Anthropic has raised eyebrows via a campaign of buying up pre-2022 books to scan for training data, while Google was <a href="https://www.404media.co/google-is-quietly-buying-code-from-play-store-developers-to-train-ai/" target="_blank">reportedly trying to buy code</a> from Google Play Store app developers. </p><p><a href="https://www.theverge.com/ai-artificial-intelligence/913806/this-startup-sells-dead-company-data-for-ai-training" target="_blank">SimpleClosure helps failed startups</a> sell their old code, chats, and emails to AI companies, and <a href="https://www.theguardian.com/technology/2026/jun/24/meta-pauses-employee-tracker-for-ai-training-amid-privacy-concerns" target="_blank">Meta had to pause</a> plans to track its own employees for AI training purposes amid internal backlash. </p><p>All of this is to avoid issues like data exhaustion or hitting a data wall, in which they lack enough good quality data to build bigger models, and to <a href="https://www.itpro.com/technology/artificial-intelligence/what-is-model-collapse-and-why-is-it-a-risk-for-enterprise-ai">avoid model collapse</a>, when training data isn't of good enough quality — or is full of AI generated content. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/google-just-spent-usd10-million-in-an-auction-for-spirit-airlines-data-100-million-emails-500-million-microsoft-teams-chats-and-30-million-lines-of-code-will-be-used-to-improve-ai-models-and-products</link>
                                                                            <description>
                            <![CDATA[ AI developers are struggling to find enough data to train their models, sparking a bidding war for failed Spirit Airlines deidentified data ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">8EkhNFMTsYQEQSmuaancDY</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/ETyDoyKVMQgNycXiNPGyjn-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 18 Aug 2026 13:46:30 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:39 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/ETyDoyKVMQgNycXiNPGyjn-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[An overhead view of a Spirit Airlines Airbus A321 on the tarmac at Los Angeles International Airport (LAX) in the United States.]]></media:description>                                                            <media:text><![CDATA[An overhead view of a Spirit Airlines Airbus A321 on the tarmac at Los Angeles International Airport (LAX) in the United States.]]></media:text>
                                <media:title type="plain"><![CDATA[An overhead view of a Spirit Airlines Airbus A321 on the tarmac at Los Angeles International Airport (LAX) in the United States.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/ETyDoyKVMQgNycXiNPGyjn-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p><a href="https://www.bbc.co.uk/news/articles/cqxlnrqjvzyo" target="_blank">Spirit Airlines collapsed</a> in May this year, but its data is set for a new lease of life after Google acquired it in an auction. </p><p>The tech giant recently outbid AI company Mercor to acquire the de-identified data, according to <a href="https://document.epiq11.com/document/getdocumentbycode?docId=4606206&amp;projectCode=SPJ&amp;source=DM&amp;utm_s" target="_blank">court documents</a>. Google bid $10 million, versus Mercor's $7.5 million, though reports note the bidding started around the $5 million mark. </p><p>A third bidder also took part but was not named. </p><p>A spokesperson for Google told <em>ITPro </em>the aim behind acquiring the defunct airline’s data lies in its potential use for AI training and product refinement. </p><p>"We acquired part of an enterprise dataset from Spirit Airlines, which can be helpful in improving our products and AI models,” the spokesperson said. </p><p>That same sentiment was highlighted by a spokesperson for Mercor, per reports from <a href="https://www.businessinsider.com/google-buys-spirit-airlines-data-ai-model-development-2026-8" target="_blank"><em>Business Insider</em></a>. </p><p>"Companies are sitting on decades of records that show how real work gets done, and that data is now some of the most valuable material for training and evaluating AI," the spokesperson said. </p><p>"We partner with leading companies to license their operational data to the labs building the next generation of models. Spirit was that same process applied to a bankruptcy estate." </p><h2 id="what-google-bought">What Google bought </h2><p>The data bought by Google includes all productivity and collaboration data, which the filing notes includes everything from emails to chats, spreadsheets to calendars, as well as "core business systems and business applications data". </p><p>This includes data on employee behavior and productivity, aircraft operations, and inventory, among other topics. Beyond that, Google has purchased workflow and process data from departments such as HR and marketing. </p><p>In total, Google is picking up 100 million emails, 500 million Microsoft Teams chats, 30 million lines of code – but not records of loyalty program members or other customer data. </p><p>Indeed, the data set does not include any consumer data or personal data, the filing noted, or any other information that could fall under data protection laws. The buyer was required to agree to not attempt to re-identify the data and link it to any person. </p><p>The buyer is allowed to transfer the data set to third parties, however. </p><p>"We will not receive any personal information from this dataset," the Google spokesperson added. "Any data we receive will be rigorously scrubbed of any personally identifiable information by a third party before receipt."</p><p>Alongside the data, Google's $10 million also buys it all of Spirit's internally developed software and applications, including all the <a href="https://www.itpro.com/software/open-source/what-red-hats-source-code-restrictions-mean-for-businesses">source code</a>, plugins, data files, libraries, <a href="https://www.itpro.com/tag/application-programming-interface">APIs</a> and documentation. </p><h2 id="the-battle-for-data">The battle for data</h2><p>The battle for the business data reveals how desperate AI developers are for fresh material to feed AI, in particular for training models, though Google didn't specifically say that was the purpose of the data set. </p><p>Anthropic has raised eyebrows via a campaign of buying up pre-2022 books to scan for training data, while Google was <a href="https://www.404media.co/google-is-quietly-buying-code-from-play-store-developers-to-train-ai/" target="_blank">reportedly trying to buy code</a> from Google Play Store app developers. </p><p><a href="https://www.theverge.com/ai-artificial-intelligence/913806/this-startup-sells-dead-company-data-for-ai-training" target="_blank">SimpleClosure helps failed startups</a> sell their old code, chats, and emails to AI companies, and <a href="https://www.theguardian.com/technology/2026/jun/24/meta-pauses-employee-tracker-for-ai-training-amid-privacy-concerns" target="_blank">Meta had to pause</a> plans to track its own employees for AI training purposes amid internal backlash. </p><p>All of this is to avoid issues like data exhaustion or hitting a data wall, in which they lack enough good quality data to build bigger models, and to <a href="https://www.itpro.com/technology/artificial-intelligence/what-is-model-collapse-and-why-is-it-a-risk-for-enterprise-ai">avoid model collapse</a>, when training data isn't of good enough quality — or is full of AI generated content. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ OpenAI forges closer ties with IBM in enterprise push ]]></title>
                                                                                                <dc:content><![CDATA[ <p>IBM and OpenAI have partnered up to help customers deploy AI at scale for core business operations.</p><p>Under the deal, IBM will create a dedicated OpenAI Practice, training thousands of consultants and engineers for expert-level certifications through the OpenAI Partner Network.</p><p>OpenAI models and products will be combined with IBM Consulting technology and expertise, the duo said, with the first areas to be targeted including financial services, government, telecommunications, and financial services.</p><p>“The organizations pulling ahead with AI are the ones turning it into a trusted part of how their business operates,” said Denise Dresser, former chief revenue officer at OpenAI. </p><p>“IBM Consulting and OpenAI are helping organizations make that shift, combining deep transformation expertise to deploy AI that is secure, operational, and aligned with real business priorities.” </p><h2 id="driving-adoption-rates">Driving adoption rates</h2><p>According to the two firms, the partnership will focus on applying AI to specific business industry processes, while also modernizing applications, redesigning back-office workflows, and strengthening <a href="https://www.itpro.com/security/28133/what-is-cyber-security">cybersecurity</a>.</p><p>"We’re starting with the business use case,” said Michael Healy, managing partner of offerings, assets and generative AI at IBM Consulting. “Then we bring IBM’s process transformation and industry expertise together with OpenAI’s models to change how that work actually gets done.”</p><p>OpenAI frontier models like GPT-5.6 and products such as <a href="https://www.itpro.com/technology/artificial-intelligence/openais-codex-app-is-now-available-on-macos-and-its-free-for-some-chatgpt-users-for-a-limited-time">Codex </a>and ChatGPT Work will be embedded into IBM’s AI platform for delivering consulting services, IBM Consulting Advantage.</p><p>This will allow operating procedures and workflows to be analyzed to identify inefficiencies and help teams automate and streamline work using AI across finance, procurement, customer operations, and HR. </p><h2 id="tech-modernization">Tech modernization</h2><p>As part of the deal, IBM said the duo also plan to help clients <a href="https://www.itpro.com/business/digital-transformation/it-leaders-are-throwing-money-away-with-legacy-systems-enterprises-report-usd370-million-in-losses-each-year-due-to-outdated-tech">modernize legacy applications</a> and <a href="https://www.itpro.com/technology/artificial-intelligence/the-pros-and-cons-of-ai-coding-in-the-it-industry">accelerate software development</a> by bringing together OpenAI Codex and ChatGPT Work with IBM’s industry, technology and domain expertise, including integration. </p><p>Here, the aim is to help organizations simplify engineering processes and speed up the delivery of new digital products and services.</p><p>“Codex as a coding assistant will accelerate the application development modernization cycle in combination with IBM Consulting Advantage’s coding harness and OpenAI’s models,” said Healy.</p><p>Following IBM’s participation in the OpenAI Daybreak Cyber Partner Program, the duo will expand their collaboration on cybersecurity. </p><p>This will involve combining OpenAI frontier AI capabilities with IBM Autonomous Security. This is an agentic AI service designed to deliver coordinated decision-making, response, and intelligence for cyber professionals. </p><p>“While enterprises are rapidly investing in AI, they are looking for practical ways to apply it across their core operations to deliver measurable business outcomes and create new commercial models,” said Andy Baldwin, global senior vice president, IBM Consulting. </p><p>“The challenge is not access to AI technologies — it’s integrating AI securely and at scale into complex enterprise environments and workflows. By embedding OpenAI’s technology with IBM Consulting’s AI assets, industry solutions, and cybersecurity capabilities, we can help clients accelerate secure, AI deployments at scale.”</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/openai-forges-closer-ties-with-ibm-in-enterprise-push</link>
                                                                            <description>
                            <![CDATA[ The duo will combine OpenAI models and products with IBM Consulting expertise ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">KREkr97iKt86MeDp9nbL27</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/t9VeuX9MMJ2baxoKmRyR2m-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 14 Aug 2026 10:23:16 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:55 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Emma Woollacott ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/aWfskavxoVSMDy6cDWtYmJ-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/t9VeuX9MMJ2baxoKmRyR2m-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[IBM logo illuminated in white against a black background at the company&#039;s exhibitor stall at Mobile World Congress 2025. ]]></media:description>                                                            <media:text><![CDATA[IBM logo illuminated in white against a black background at the company&#039;s exhibitor stall at Mobile World Congress 2025. ]]></media:text>
                                <media:title type="plain"><![CDATA[IBM logo illuminated in white against a black background at the company&#039;s exhibitor stall at Mobile World Congress 2025. ]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/t9VeuX9MMJ2baxoKmRyR2m-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>IBM and OpenAI have partnered up to help customers deploy AI at scale for core business operations.</p><p>Under the deal, IBM will create a dedicated OpenAI Practice, training thousands of consultants and engineers for expert-level certifications through the OpenAI Partner Network.</p><p>OpenAI models and products will be combined with IBM Consulting technology and expertise, the duo said, with the first areas to be targeted including financial services, government, telecommunications, and financial services.</p><p>“The organizations pulling ahead with AI are the ones turning it into a trusted part of how their business operates,” said Denise Dresser, former chief revenue officer at OpenAI. </p><p>“IBM Consulting and OpenAI are helping organizations make that shift, combining deep transformation expertise to deploy AI that is secure, operational, and aligned with real business priorities.” </p><h2 id="driving-adoption-rates">Driving adoption rates</h2><p>According to the two firms, the partnership will focus on applying AI to specific business industry processes, while also modernizing applications, redesigning back-office workflows, and strengthening <a href="https://www.itpro.com/security/28133/what-is-cyber-security">cybersecurity</a>.</p><p>"We’re starting with the business use case,” said Michael Healy, managing partner of offerings, assets and generative AI at IBM Consulting. “Then we bring IBM’s process transformation and industry expertise together with OpenAI’s models to change how that work actually gets done.”</p><p>OpenAI frontier models like GPT-5.6 and products such as <a href="https://www.itpro.com/technology/artificial-intelligence/openais-codex-app-is-now-available-on-macos-and-its-free-for-some-chatgpt-users-for-a-limited-time">Codex </a>and ChatGPT Work will be embedded into IBM’s AI platform for delivering consulting services, IBM Consulting Advantage.</p><p>This will allow operating procedures and workflows to be analyzed to identify inefficiencies and help teams automate and streamline work using AI across finance, procurement, customer operations, and HR. </p><h2 id="tech-modernization">Tech modernization</h2><p>As part of the deal, IBM said the duo also plan to help clients <a href="https://www.itpro.com/business/digital-transformation/it-leaders-are-throwing-money-away-with-legacy-systems-enterprises-report-usd370-million-in-losses-each-year-due-to-outdated-tech">modernize legacy applications</a> and <a href="https://www.itpro.com/technology/artificial-intelligence/the-pros-and-cons-of-ai-coding-in-the-it-industry">accelerate software development</a> by bringing together OpenAI Codex and ChatGPT Work with IBM’s industry, technology and domain expertise, including integration. </p><p>Here, the aim is to help organizations simplify engineering processes and speed up the delivery of new digital products and services.</p><p>“Codex as a coding assistant will accelerate the application development modernization cycle in combination with IBM Consulting Advantage’s coding harness and OpenAI’s models,” said Healy.</p><p>Following IBM’s participation in the OpenAI Daybreak Cyber Partner Program, the duo will expand their collaboration on cybersecurity. </p><p>This will involve combining OpenAI frontier AI capabilities with IBM Autonomous Security. This is an agentic AI service designed to deliver coordinated decision-making, response, and intelligence for cyber professionals. </p><p>“While enterprises are rapidly investing in AI, they are looking for practical ways to apply it across their core operations to deliver measurable business outcomes and create new commercial models,” said Andy Baldwin, global senior vice president, IBM Consulting. </p><p>“The challenge is not access to AI technologies — it’s integrating AI securely and at scale into complex enterprise environments and workflows. By embedding OpenAI’s technology with IBM Consulting’s AI assets, industry solutions, and cybersecurity capabilities, we can help clients accelerate secure, AI deployments at scale.”</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ This new technique could improve AI output accuracy by 80%  ]]></title>
                                                                                                <dc:content><![CDATA[ <p>A new data refinement technique could help improve the accuracy of AI outputs and help build trust in the technology, according to researchers. </p><p>The UK’s National Innovation Centre for Data (NICD) at Newcastle University has developed a technique that can boost accuracy by as much as 80% – a major improvement, although AI answers still remain imperfect. </p><p>Researchers <a href="https://arxiv.org/abs/2606.05901"><u>achieved this</u></a> by combining large language models (LLMs) with graph-based knowledge databases, a storage method which creates networks of data entities to provide deeper semantic meaning and context. </p><p>This combination could help reduce “hallucinations” and omissions in AI outputs. </p><p>"The work of AI experts at the UK National Innovation Centre for Data has shown that integration with a graph database can significantly reduce the two most significant problems holding back the exploitation of LLMs for real-world applications: hallucinations and omissions," said Paul Watson, the director of the NICD. </p><p>"This is especially important for organizations deploying LLMs in applications where regulatory compliance and the avoidance of reputational or financial damage is key." </p><h2 id="how-it-works">How it works</h2><p>Select models were tested against 510 questions designed to check reasoning across different sources. Combining an AI model with graph database technology led to the best results, significantly outperforming AI models that rely on their training data as well as <a href="https://www.itpro.com/technology/artificial-intelligence/what-is-retrieval-augmented-generation-rag">retrieval augmentation generation (RAG)</a> approaches. </p><p>The researchers' technique, which combined RAG with a simple graph system, increased the rate of accurate answers from 29% to 66%.</p><p>A knowledge graph is a structured set of information that can be labelled and highlights the relationship between ideas, and could provide vital in helping to improve context for AI models. </p><p>"Rather than operating over unstructured, chunked text documents, a knowledge graph (KG) is either harnessed directly or created using an LLM from a set of related documents," the researchers explain in their paper.</p><p>Challenges remain, however, most notably in creating knowledge graphs that work well with AI. Too complex, and they risk filling an AI's context window; too simple, and though "easily digestible" by an AI, the results aren't as good. </p><p>Still, researchers said the combined RAG-knowledge graph approach was a "promising direction" for improving AI accuracy. Plus, that approach uses fewer tokens for better quality results, they noted. </p><h2 id="hallucination-risks">Hallucination risks</h2><p>Improving the accuracy of AI will help reduce the significant risks to businesses, noted Dr Jim Webber, the chief scientist at Neo4j. </p><p>"In an agentic world, where autonomous systems can make significant decisions over time, this cannot hold," he said. </p><p>Webber noted that the work comes with another benefit: cost reduction. Improving accuracy in AI normally involves larger models with more data, which raises costs significantly – a fact seen in recent price increases. </p><p>Notably, the NICD's technique leads to better results at lower costs and "shows that better models alone cannot solve the problem, but that the right data at the right time can provide significant benefits in terms of accuracy, responsiveness, and cost."<em> </em></p><p>"They have shown that enterprises no longer need to compromise on the quality of their AI, while they can compromise on cost. This is an unusual and highly welcome finding,” he added. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/business/data-and-insights/this-new-technique-could-improve-ai-output-accuracy-by-80-percent-and-tackle-hallucinations-once-and-for-all</link>
                                                                            <description>
                            <![CDATA[ Researchers believe the new method will improve accuracy and help reduce costs ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">G8mxae6jDqQfVYCpYCN4xN</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/eULhhy5kURKAAtprURE8GW-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 04 Aug 2026 10:38:44 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:45 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Data and Insights]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                    <category><![CDATA[Business]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/eULhhy5kURKAAtprURE8GW-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[AI training concept image showing a robotic hand with human silhouette standing on finger tip while looking at digital interfaces.]]></media:description>                                                            <media:text><![CDATA[AI training concept image showing a robotic hand with human silhouette standing on finger tip while looking at digital interfaces.]]></media:text>
                                <media:title type="plain"><![CDATA[AI training concept image showing a robotic hand with human silhouette standing on finger tip while looking at digital interfaces.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/eULhhy5kURKAAtprURE8GW-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>A new data refinement technique could help improve the accuracy of AI outputs and help build trust in the technology, according to researchers. </p><p>The UK’s National Innovation Centre for Data (NICD) at Newcastle University has developed a technique that can boost accuracy by as much as 80% – a major improvement, although AI answers still remain imperfect. </p><p>Researchers <a href="https://arxiv.org/abs/2606.05901"><u>achieved this</u></a> by combining large language models (LLMs) with graph-based knowledge databases, a storage method which creates networks of data entities to provide deeper semantic meaning and context. </p><p>This combination could help reduce “hallucinations” and omissions in AI outputs. </p><p>"The work of AI experts at the UK National Innovation Centre for Data has shown that integration with a graph database can significantly reduce the two most significant problems holding back the exploitation of LLMs for real-world applications: hallucinations and omissions," said Paul Watson, the director of the NICD. </p><p>"This is especially important for organizations deploying LLMs in applications where regulatory compliance and the avoidance of reputational or financial damage is key." </p><h2 id="how-it-works">How it works</h2><p>Select models were tested against 510 questions designed to check reasoning across different sources. Combining an AI model with graph database technology led to the best results, significantly outperforming AI models that rely on their training data as well as <a href="https://www.itpro.com/technology/artificial-intelligence/what-is-retrieval-augmented-generation-rag">retrieval augmentation generation (RAG)</a> approaches. </p><p>The researchers' technique, which combined RAG with a simple graph system, increased the rate of accurate answers from 29% to 66%.</p><p>A knowledge graph is a structured set of information that can be labelled and highlights the relationship between ideas, and could provide vital in helping to improve context for AI models. </p><p>"Rather than operating over unstructured, chunked text documents, a knowledge graph (KG) is either harnessed directly or created using an LLM from a set of related documents," the researchers explain in their paper.</p><p>Challenges remain, however, most notably in creating knowledge graphs that work well with AI. Too complex, and they risk filling an AI's context window; too simple, and though "easily digestible" by an AI, the results aren't as good. </p><p>Still, researchers said the combined RAG-knowledge graph approach was a "promising direction" for improving AI accuracy. Plus, that approach uses fewer tokens for better quality results, they noted. </p><h2 id="hallucination-risks">Hallucination risks</h2><p>Improving the accuracy of AI will help reduce the significant risks to businesses, noted Dr Jim Webber, the chief scientist at Neo4j. </p><p>"In an agentic world, where autonomous systems can make significant decisions over time, this cannot hold," he said. </p><p>Webber noted that the work comes with another benefit: cost reduction. Improving accuracy in AI normally involves larger models with more data, which raises costs significantly – a fact seen in recent price increases. </p><p>Notably, the NICD's technique leads to better results at lower costs and "shows that better models alone cannot solve the problem, but that the right data at the right time can provide significant benefits in terms of accuracy, responsiveness, and cost."<em> </em></p><p>"They have shown that enterprises no longer need to compromise on the quality of their AI, while they can compromise on cost. This is an unusual and highly welcome finding,” he added. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Oracle integrates Google's Gemini AI models into enterprise apps ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Oracle is expanding its use of Google's AI models across its enterprise software portfolio.</p><p>The company is adding Google's Gemini models to its AI Agent Studio for Fusion Applications, a development platform that enables organizations to build, connect, execute, and run AI automation and agentic applications using reusable Oracle, partner, and external agents. </p><p>It also plans to use Gemini models for embedded AI use cases in Oracle Fusion Applications and Oracle NetSuite.</p><p>“Organizations around the world trust Google Cloud’s full AI stack to power critical enterprise workflows and agents,” said Satish Thomas, vice president of Google Cloud. </p><p>“Our expanded partnership with Oracle is designed to make it easier for organizations to use Gemini in the applications and agentic workflows they rely on to automate workflows, accelerate decisions, and drive outcomes.”</p><p>The deal builds on customers’ existing access to Gemini models through Oracle Cloud Infrastructure (OCI) Enterprise AI through integration with Gemini Enterprise Agent Platform. </p><p>“Our partnership with Oracle brings Google’s most capable AI models directly into the core application workflows global businesses rely on every day," said Kevin Ichhpurani, president of global partner ecosystem at Google Cloud. </p><p>“Together, we are making it seamless for enterprises to apply powerful and cost-efficient AI directly where business decisions happen.”</p><p>Oracle said customers and partners will get more choice when building Fusion-native agents and agentic applications, along with expanded multi-modal capabilities. They'll be able, for example, to access Gemini 3.1 Flash Lite, a high-efficiency model engineered for optimal price-performance and Gemini 3.5 Flash for more complex reasoning and specialized tasks, including video and presentation creation.</p><p>“To achieve the best business outcomes, organizations need the flexibility to choose the AI model best suited to each problem,” said Chris Leone, EVP of applications development at Oracle. </p><p>“By bringing Gemini to Oracle AI Agent Studio for Fusion Applications, we are giving customers and partners greater choice as they build and extend agents and agentic applications that reason through complex, real-world business challenges. Oracle Fusion Applications then turn that reasoning into action through governed workflows, approvals, and transactions.”</p><p>Oracle also plans to use Gemini models for embedded AI use cases in Netsuite and Oracle Fusion Applications, which include Oracle Fusion Cloud Enterprise Resource Planning (ERP), Oracle Fusion Cloud Human Capital Management (HCM), Oracle Fusion Cloud Supply Chain &amp; Manufacturing (SCM) and Oracle Fusion Cloud Customer Experience (CX).</p><p>In each case, it will choose the model that can deliver the best price-performance for specific customer scenarios.</p><p>“AI is at the core of how customers use and experience NetSuite and choosing the right model for the right use case is critical to helping them get more value from AI,” said Evan Goldberg, founder and executive vice president, Oracle NetSuite. </p><p>“As we evaluate various AI use cases in NetSuite, we are working with leading large language models, like Google’s Gemini, to help customers improve visibility, automate work, and move from insight to action within NetSuite.”</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/oracle-integrates-googles-gemini-ai-models-into-enterprise-apps</link>
                                                                            <description>
                            <![CDATA[ The deal aims to give broader access to Gemini models that can support AI agents and accelerate development ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">WCRRsD4nbaxyXgYc2sVj4C</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/CLcrZfFWca5tnSjCbNuQ7C-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 31 Jul 2026 10:55:03 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:40 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Emma Woollacott ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/aWfskavxoVSMDy6cDWtYmJ-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/CLcrZfFWca5tnSjCbNuQ7C-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Google Gemini Pro logo displayed on a smartphone with &#039;AI&#039; lettering shown in background]]></media:description>                                                            <media:text><![CDATA[Google Gemini Pro logo displayed on a smartphone with &#039;AI&#039; lettering shown in background]]></media:text>
                                <media:title type="plain"><![CDATA[Google Gemini Pro logo displayed on a smartphone with &#039;AI&#039; lettering shown in background]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/CLcrZfFWca5tnSjCbNuQ7C-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Oracle is expanding its use of Google's AI models across its enterprise software portfolio.</p><p>The company is adding Google's Gemini models to its AI Agent Studio for Fusion Applications, a development platform that enables organizations to build, connect, execute, and run AI automation and agentic applications using reusable Oracle, partner, and external agents. </p><p>It also plans to use Gemini models for embedded AI use cases in Oracle Fusion Applications and Oracle NetSuite.</p><p>“Organizations around the world trust Google Cloud’s full AI stack to power critical enterprise workflows and agents,” said Satish Thomas, vice president of Google Cloud. </p><p>“Our expanded partnership with Oracle is designed to make it easier for organizations to use Gemini in the applications and agentic workflows they rely on to automate workflows, accelerate decisions, and drive outcomes.”</p><p>The deal builds on customers’ existing access to Gemini models through Oracle Cloud Infrastructure (OCI) Enterprise AI through integration with Gemini Enterprise Agent Platform. </p><p>“Our partnership with Oracle brings Google’s most capable AI models directly into the core application workflows global businesses rely on every day," said Kevin Ichhpurani, president of global partner ecosystem at Google Cloud. </p><p>“Together, we are making it seamless for enterprises to apply powerful and cost-efficient AI directly where business decisions happen.”</p><p>Oracle said customers and partners will get more choice when building Fusion-native agents and agentic applications, along with expanded multi-modal capabilities. They'll be able, for example, to access Gemini 3.1 Flash Lite, a high-efficiency model engineered for optimal price-performance and Gemini 3.5 Flash for more complex reasoning and specialized tasks, including video and presentation creation.</p><p>“To achieve the best business outcomes, organizations need the flexibility to choose the AI model best suited to each problem,” said Chris Leone, EVP of applications development at Oracle. </p><p>“By bringing Gemini to Oracle AI Agent Studio for Fusion Applications, we are giving customers and partners greater choice as they build and extend agents and agentic applications that reason through complex, real-world business challenges. Oracle Fusion Applications then turn that reasoning into action through governed workflows, approvals, and transactions.”</p><p>Oracle also plans to use Gemini models for embedded AI use cases in Netsuite and Oracle Fusion Applications, which include Oracle Fusion Cloud Enterprise Resource Planning (ERP), Oracle Fusion Cloud Human Capital Management (HCM), Oracle Fusion Cloud Supply Chain &amp; Manufacturing (SCM) and Oracle Fusion Cloud Customer Experience (CX).</p><p>In each case, it will choose the model that can deliver the best price-performance for specific customer scenarios.</p><p>“AI is at the core of how customers use and experience NetSuite and choosing the right model for the right use case is critical to helping them get more value from AI,” said Evan Goldberg, founder and executive vice president, Oracle NetSuite. </p><p>“As we evaluate various AI use cases in NetSuite, we are working with leading large language models, like Google’s Gemini, to help customers improve visibility, automate work, and move from insight to action within NetSuite.”</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Big tech faces an adapt or die predicament with open weight AI models ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Open-weight AI models have rapidly emerged as the latest flashpoint in big tech, with popular Chinese models posing a serious threat to leading walled garden wardens. </p><p>Fuelled by an undercurrent of geopolitical jostling between the US and China, models released by the latter’s burgeoning AI industry have had tech giants in a state of red alert in recent weeks. </p><p>Kimi K3, a Chinese model developed by Moonshot AI, has wowed users and appears more than capable of going toe-to-toe with some of the best US-made options on the market right now. </p><p>Competing with powerful Chinese models isn’t a new experience for the US-dominated AI market, with DeepSeek rocking the market in 2025. What appears to be troubling them – and indeed lawmakers in Washington DC –  this time is the fact Kimi K3 is an open-weight model. This means its training parameters (or weights) are publicly available for anyone to download, modify, and run. </p><p>That poses a direct threat to major providers intent on keeping enterprise users locked into their own ecosystems. Jeff Watkins, chief AI officer of Leeds-based AI consultancy, NorthStar Intelligence, said these options could offer a more flexible and economically viable alternative to closed-source providers. </p><p>This is because enterprises can run open-weight models on their own infrastructure, using their own data. They are also easily customizable, enabling users to tweak them based on changing needs and, crucially, without relying on a single provider.</p><p>“Open-weight models are becoming increasingly attractive because they give organizations far more control than closed commercial APIs,” Watkins told <em>ITPro</em>. </p><p>“For enterprises, the biggest advantages of open-weight models are control over deployment, upgrades, hosting, data residency, access controls and long-term operating costs.”</p><p>“These characteristics make open-weight models particularly attractive for regulated industries where resilience, compliance and predictable operating costs are critical.”</p><h2 id="open-source-vs-open-weight">Open source vs open-weight</h2><p>Open source AI models have become equally attractive to enterprises in recent years, partly for the same reasons: control, flexibility, and independence. Research published in November 2025 found <a href="https://www.itpro.com/software/open-source/open-source-ai-performance-cost-savings-proprietary-models-linux-foundation"><u>open source AI models perform on-par with closed source options</u></a> and are typically cheaper. </p><p>With the limelight on open-weight options, it’s important to make a clear distinction between them. While there are similarities, open-weight models don’t include the underlying training data used to build them. </p><p>This does have benefits though, according to OpenUK CEO Amanda Brock, enabling enterprises to build highly customized models and “giving access to innovation”. </p><p>“For innovators, accompanying this with the right documentation enables them to rebuild it into their own model on their own data,” she told <em>ITPro</em>. </p><p>“We saw this happen on Hugging Face last year with DeepSeek R1, where the community built Open R1 rather than have to rely on a single model provider,” Brock added. “This is a great example of innovators iterating in the tradition of open source. It also enables products to be built inexpensively for end users, and to give access to all.”</p><h2 id="betting-big">Betting big</h2><p>The sheer volume of open-weight and open source models now on the market does highlight a growing shift, according to Brock. </p><p>The fact that some of the leading models are Chinese-made is equally important, as the country has made a conscious effort to compete with US providers by creating a level playing field. </p><p>“China took the decision to shift to a clear open source strategy about eight years ago,” Brock noted. “They’d seen how software has evolved and the US’s position in it.”</p><p>“Big tech has been enabled by adopting and using open source, where the software becomes a de-facto standard and those leading in it are central to the ecosystem. A model of open source that’s about big tech collaboration saving costs and building out standards lower in the software stack evolved, and open source offers adoption at a scale that closed proprietary software cannot compete with.”</p><h2 id="the-stable-s-open-and-the-horse-has-bolted">The stable’s open and the horse has bolted</h2><p>Watkins echoed Brock’s comments, noting that China’s focus on open-weight AI development “makes strategic sense” and will undoubtedly put pressure on US providers. </p><p>Presenting these as viable alternatives fundamentally undermines the image of the US as the go-to marketplace for AI. </p><p>“Rather than competing solely through proprietary hosted services, releasing capable open-weight models encourages global adoption, builds developer ecosystems, and creates competitive pressure on US providers,” he told <em>ITPro</em>. </p><p>There are signs that this pressure is mounting given recent speculation about a pushback by US authorities. <a href="https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi"><u>Reports from </u><u><em>Axios</em></u></a><em> </em>this month suggested the White House could consider imposing tight conditions, or even restrictions, on US firms working with these models. </p><p>These potential moves have a geopolitical motive, but regardless, it’s clear the horse has bolted at this stage. </p><p>A host of big tech companies including Microsoft, Nvidia, IBM and more cautioned against “premature restrictions” in an <a href="https://images.nvidia.com/pdf/Open-Weights-and-American-AI-Leadership.pdf"><u>open letter</u></a> last week. Nvidia CEO Jensen Huang even made an X account <a href="https://x.com/JensenHuang/status/2080643682408321103?s=20"><u>to get his message across</u></a>. </p><p>This was followed by another <a href="https://blogs.nvidia.com/blog/open-secure-ai-alliance/"><u>pan-industry call to support</u></a> open-weight models in response to the OpenAI-Hugging Face incident, in which the latter was forced<a href="https://www.itpro.com/security/an-unprecedented-cyber-incident-how-openai-models-breached-hugging-face-and-why-it-could-herald-a-new-phase-of-ai-powered-cyber-crime"><u> to use a Chinese-made model</u></a> to stop the hack. </p><p>There is a certain irony that a security incident involving a leading US AI provider has spurred on industry-wide support for the same models that are spooking US authorities. </p><p>Watkins suggested that the US AI industry now faces an uncomfortable truth. It may have led the generative AI boom, but it now has to compete with an ecosystem of models designed to undercut it. </p><p>“Chinese providers have demonstrated that frontier-quality AI is no longer exclusively a US capability,” he told <em>ITPro</em>.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/big-tech-faces-an-adapt-or-die-predicament-with-open-weight-ai-models</link>
                                                                            <description>
                            <![CDATA[ An array of US tech providers are now defending the AI models that undermine the illusion of big tech exceptionalism ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">6S4UWGbEQiCp677Njs5ruB</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/bLSsYgswXfVWXGiXgSJvj8-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 30 Jul 2026 09:47:47 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:50 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/bLSsYgswXfVWXGiXgSJvj8-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Conceptual image of a large-scale futuristic data center on a peach grid showcasing a prominent glowing AI cube cluster emerging from an orange container, with extensive colorful wiring linking to surrounding illuminated server racks in precise formations, representing advanced artificial intelligence infrastructure, machine learning networks, and interconnected digital systems in a high-tech conceptual 3D scene with vibrant effects.]]></media:description>                                                            <media:text><![CDATA[Conceptual image of a large-scale futuristic data center on a peach grid showcasing a prominent glowing AI cube cluster emerging from an orange container, with extensive colorful wiring linking to surrounding illuminated server racks in precise formations, representing advanced artificial intelligence infrastructure, machine learning networks, and interconnected digital systems in a high-tech conceptual 3D scene with vibrant effects.]]></media:text>
                                <media:title type="plain"><![CDATA[Conceptual image of a large-scale futuristic data center on a peach grid showcasing a prominent glowing AI cube cluster emerging from an orange container, with extensive colorful wiring linking to surrounding illuminated server racks in precise formations, representing advanced artificial intelligence infrastructure, machine learning networks, and interconnected digital systems in a high-tech conceptual 3D scene with vibrant effects.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/bLSsYgswXfVWXGiXgSJvj8-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Open-weight AI models have rapidly emerged as the latest flashpoint in big tech, with popular Chinese models posing a serious threat to leading walled garden wardens. </p><p>Fuelled by an undercurrent of geopolitical jostling between the US and China, models released by the latter’s burgeoning AI industry have had tech giants in a state of red alert in recent weeks. </p><p>Kimi K3, a Chinese model developed by Moonshot AI, has wowed users and appears more than capable of going toe-to-toe with some of the best US-made options on the market right now. </p><p>Competing with powerful Chinese models isn’t a new experience for the US-dominated AI market, with DeepSeek rocking the market in 2025. What appears to be troubling them – and indeed lawmakers in Washington DC –  this time is the fact Kimi K3 is an open-weight model. This means its training parameters (or weights) are publicly available for anyone to download, modify, and run. </p><p>That poses a direct threat to major providers intent on keeping enterprise users locked into their own ecosystems. Jeff Watkins, chief AI officer of Leeds-based AI consultancy, NorthStar Intelligence, said these options could offer a more flexible and economically viable alternative to closed-source providers. </p><p>This is because enterprises can run open-weight models on their own infrastructure, using their own data. They are also easily customizable, enabling users to tweak them based on changing needs and, crucially, without relying on a single provider.</p><p>“Open-weight models are becoming increasingly attractive because they give organizations far more control than closed commercial APIs,” Watkins told <em>ITPro</em>. </p><p>“For enterprises, the biggest advantages of open-weight models are control over deployment, upgrades, hosting, data residency, access controls and long-term operating costs.”</p><p>“These characteristics make open-weight models particularly attractive for regulated industries where resilience, compliance and predictable operating costs are critical.”</p><h2 id="open-source-vs-open-weight">Open source vs open-weight</h2><p>Open source AI models have become equally attractive to enterprises in recent years, partly for the same reasons: control, flexibility, and independence. Research published in November 2025 found <a href="https://www.itpro.com/software/open-source/open-source-ai-performance-cost-savings-proprietary-models-linux-foundation"><u>open source AI models perform on-par with closed source options</u></a> and are typically cheaper. </p><p>With the limelight on open-weight options, it’s important to make a clear distinction between them. While there are similarities, open-weight models don’t include the underlying training data used to build them. </p><p>This does have benefits though, according to OpenUK CEO Amanda Brock, enabling enterprises to build highly customized models and “giving access to innovation”. </p><p>“For innovators, accompanying this with the right documentation enables them to rebuild it into their own model on their own data,” she told <em>ITPro</em>. </p><p>“We saw this happen on Hugging Face last year with DeepSeek R1, where the community built Open R1 rather than have to rely on a single model provider,” Brock added. “This is a great example of innovators iterating in the tradition of open source. It also enables products to be built inexpensively for end users, and to give access to all.”</p><h2 id="betting-big">Betting big</h2><p>The sheer volume of open-weight and open source models now on the market does highlight a growing shift, according to Brock. </p><p>The fact that some of the leading models are Chinese-made is equally important, as the country has made a conscious effort to compete with US providers by creating a level playing field. </p><p>“China took the decision to shift to a clear open source strategy about eight years ago,” Brock noted. “They’d seen how software has evolved and the US’s position in it.”</p><p>“Big tech has been enabled by adopting and using open source, where the software becomes a de-facto standard and those leading in it are central to the ecosystem. A model of open source that’s about big tech collaboration saving costs and building out standards lower in the software stack evolved, and open source offers adoption at a scale that closed proprietary software cannot compete with.”</p><h2 id="the-stable-s-open-and-the-horse-has-bolted">The stable’s open and the horse has bolted</h2><p>Watkins echoed Brock’s comments, noting that China’s focus on open-weight AI development “makes strategic sense” and will undoubtedly put pressure on US providers. </p><p>Presenting these as viable alternatives fundamentally undermines the image of the US as the go-to marketplace for AI. </p><p>“Rather than competing solely through proprietary hosted services, releasing capable open-weight models encourages global adoption, builds developer ecosystems, and creates competitive pressure on US providers,” he told <em>ITPro</em>. </p><p>There are signs that this pressure is mounting given recent speculation about a pushback by US authorities. <a href="https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi"><u>Reports from </u><u><em>Axios</em></u></a><em> </em>this month suggested the White House could consider imposing tight conditions, or even restrictions, on US firms working with these models. </p><p>These potential moves have a geopolitical motive, but regardless, it’s clear the horse has bolted at this stage. </p><p>A host of big tech companies including Microsoft, Nvidia, IBM and more cautioned against “premature restrictions” in an <a href="https://images.nvidia.com/pdf/Open-Weights-and-American-AI-Leadership.pdf"><u>open letter</u></a> last week. Nvidia CEO Jensen Huang even made an X account <a href="https://x.com/JensenHuang/status/2080643682408321103?s=20"><u>to get his message across</u></a>. </p><p>This was followed by another <a href="https://blogs.nvidia.com/blog/open-secure-ai-alliance/"><u>pan-industry call to support</u></a> open-weight models in response to the OpenAI-Hugging Face incident, in which the latter was forced<a href="https://www.itpro.com/security/an-unprecedented-cyber-incident-how-openai-models-breached-hugging-face-and-why-it-could-herald-a-new-phase-of-ai-powered-cyber-crime"><u> to use a Chinese-made model</u></a> to stop the hack. </p><p>There is a certain irony that a security incident involving a leading US AI provider has spurred on industry-wide support for the same models that are spooking US authorities. </p><p>Watkins suggested that the US AI industry now faces an uncomfortable truth. It may have led the generative AI boom, but it now has to compete with an ecosystem of models designed to undercut it. </p><p>“Chinese providers have demonstrated that frontier-quality AI is no longer exclusively a US capability,” he told <em>ITPro</em>.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Two-thirds of workers are so fed up with ‘AI slop’ that they ‘feel nostalgic for pre-AI work’ ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Knowledge workers are secretly yearning for the days before AI entered the mix, with three-in-ten saying they preferred work before adoption of the technology. </p><p>In a survey of office workers by digital transformation firm Adaptavist, 65% said they preferred the pre-AI era and 38% would remove generative <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a> from the world entirely if they had the chance.</p><p>Younger workers in particular are among those most frustrated by AI, with 40% of Gen Z and Millennials saying they would scrap generative AI tools, compared with 32% of Gen X and 29% of Boomers.</p><p>The findings mark a shift in feelings about AI, the researchers said. While businesses continue to invest heavily in the technology, public concern is growing over its <a href="https://www.itpro.com/security/uk-business-leaders-think-ai-will-create-more-jobs-that-it-destroys-the-reality-lies-somewhere-in-between">impact on jobs</a>, workplace surveillance, privacy, and security.</p><p>Meanwhile, AI has introduced new pressures for workers, eroded the value of skilled work, and left employees feeling less engaged and less valued than before.</p><p>"These findings point to an underlying gap we see in most AI implementations," said Neal Riley, AI innovation lead at Adaptavist. </p><p>"It is much easier for organizations to focus on adoption metrics – who is using AI, how often they are using it – than it is to measure its impact on the work itself."</p><p>Just over three-in-ten (31%) of the workers who said they'd like to get rid of AI said it was because they believed it reduces creativity. Almost as many had ethical doubts, with 29% citing concerns over misuse and 28% worrying about surveillance and privacy.</p><h2 id="ai-slop-is-a-leading-frustration">‘AI slop’ is a leading frustration</h2><p>Notably, nearly half (46%) said that dealing with low-quality 'AI slop' makes their job feel less meaningful and more repetitive. More than one-third (37%), for example, admitted it has made them less engaged at work overall.</p><p>It's not as if workers are seeing the big efficiency benefits often associated with AI, the study found. More than four-in-ten (42%) now spend more time verifying and fact-checking AI output than they actually save by using it. </p><p>Almost half said that poor quality AI-generated work is actively slowing down their projects, while 55% believe it is reducing overall team efficiency.</p><p>The findings from Adaptavist align with <a href="https://www.itpro.com/security/cyber-professionals-are-flocking-to-ai-tools-but-theyre-getting-tired-of-fixing-mistakes-and-reviewing-outputs"><u>recent research from ISC2</u></a>, which found nearly two-thirds (65%) of cyber professionals are spending more time deciding whether to trust or act on AI-generated recommendations. </p><p>A similar number (63%) said they now find themselves reviewing and validating AI outputs, creating larger workloads and essentially wasting time on a daily basis. </p><p>‘AI slop’ has become a recurring buzzword over the last 18 months. Microsoft CEO Satya Nadella <a href="https://www.itpro.com/technology/artificial-intelligence/satya-nadella-microsoft-ai-slop-2026"><u>penned a blog post in January</u></a> this year calling on the industry to disregard the term. </p><p>Yet research shows this is having a direct impact on workplace efficiency and productivity. As <a href="https://www.itpro.com/technology/artificial-intelligence/workers-are-wasting-half-a-day-each-week-fixing-ai-workslop"><u><em>ITPro </em></u><u>reported in January</u></a>, analysis from Zapier found employees are working an extra four and a half hours each week cleaning up mistakes. </p><p>The consequences of not acting on low quality AI-generated outputs can be grave, the study noted. Respondents reported having had work rejected due to poor AI outputs, while others highlighted customer complaints and even security incidents. </p><h2 id="ai-is-adding-pressure">AI is adding pressure</h2><p>Workers are also feeling pressured, with half believing that their performance is now being compared – fairly or not – to AI-generated output. </p><p>Around a quarter are facing intense pressure to improve performance, improve quality, and be more efficient, just to keep pace in a machine-accelerated environment. </p><p>Another quarter said they use AI simply to meet workload demands, and 23% rely on the technology to keep up with colleagues.</p><p>"By understanding the nature of the work and the different value streams across your business, you can more accurately measure outcomes and impact rather than simply counting actions," said Riley. </p><p>"When AI is introduced thoughtfully, with the right guardrails and genuine support for the people using it, it can enhance rather than erode what makes work meaningful and impactful."</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/business/business-strategy/two-thirds-of-workers-are-so-fed-up-with-ai-slop-that-they-feel-nostalgic-for-pre-ai-work</link>
                                                                            <description>
                            <![CDATA[ A survey has revealed that dealing with low-quality 'AI slop' is making jobs feel less meaningful and more repetitive ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">Y6iUZps9pEAodxsSxWSmgR</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/zRwQ5FX2RTJmdqXdbBAjqn-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 27 Jul 2026 09:25:08 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 15:00:17 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Emma Woollacott ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/aWfskavxoVSMDy6cDWtYmJ-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/zRwQ5FX2RTJmdqXdbBAjqn-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Male office worker sitting at a desk in a dimly lit room with hands held to his face in frustration with light from screen shining on face.]]></media:description>                                                            <media:text><![CDATA[Male office worker sitting at a desk in a dimly lit room with hands held to his face in frustration with light from screen shining on face.]]></media:text>
                                <media:title type="plain"><![CDATA[Male office worker sitting at a desk in a dimly lit room with hands held to his face in frustration with light from screen shining on face.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/zRwQ5FX2RTJmdqXdbBAjqn-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Knowledge workers are secretly yearning for the days before AI entered the mix, with three-in-ten saying they preferred work before adoption of the technology. </p><p>In a survey of office workers by digital transformation firm Adaptavist, 65% said they preferred the pre-AI era and 38% would remove generative <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a> from the world entirely if they had the chance.</p><p>Younger workers in particular are among those most frustrated by AI, with 40% of Gen Z and Millennials saying they would scrap generative AI tools, compared with 32% of Gen X and 29% of Boomers.</p><p>The findings mark a shift in feelings about AI, the researchers said. While businesses continue to invest heavily in the technology, public concern is growing over its <a href="https://www.itpro.com/security/uk-business-leaders-think-ai-will-create-more-jobs-that-it-destroys-the-reality-lies-somewhere-in-between">impact on jobs</a>, workplace surveillance, privacy, and security.</p><p>Meanwhile, AI has introduced new pressures for workers, eroded the value of skilled work, and left employees feeling less engaged and less valued than before.</p><p>"These findings point to an underlying gap we see in most AI implementations," said Neal Riley, AI innovation lead at Adaptavist. </p><p>"It is much easier for organizations to focus on adoption metrics – who is using AI, how often they are using it – than it is to measure its impact on the work itself."</p><p>Just over three-in-ten (31%) of the workers who said they'd like to get rid of AI said it was because they believed it reduces creativity. Almost as many had ethical doubts, with 29% citing concerns over misuse and 28% worrying about surveillance and privacy.</p><h2 id="ai-slop-is-a-leading-frustration">‘AI slop’ is a leading frustration</h2><p>Notably, nearly half (46%) said that dealing with low-quality 'AI slop' makes their job feel less meaningful and more repetitive. More than one-third (37%), for example, admitted it has made them less engaged at work overall.</p><p>It's not as if workers are seeing the big efficiency benefits often associated with AI, the study found. More than four-in-ten (42%) now spend more time verifying and fact-checking AI output than they actually save by using it. </p><p>Almost half said that poor quality AI-generated work is actively slowing down their projects, while 55% believe it is reducing overall team efficiency.</p><p>The findings from Adaptavist align with <a href="https://www.itpro.com/security/cyber-professionals-are-flocking-to-ai-tools-but-theyre-getting-tired-of-fixing-mistakes-and-reviewing-outputs"><u>recent research from ISC2</u></a>, which found nearly two-thirds (65%) of cyber professionals are spending more time deciding whether to trust or act on AI-generated recommendations. </p><p>A similar number (63%) said they now find themselves reviewing and validating AI outputs, creating larger workloads and essentially wasting time on a daily basis. </p><p>‘AI slop’ has become a recurring buzzword over the last 18 months. Microsoft CEO Satya Nadella <a href="https://www.itpro.com/technology/artificial-intelligence/satya-nadella-microsoft-ai-slop-2026"><u>penned a blog post in January</u></a> this year calling on the industry to disregard the term. </p><p>Yet research shows this is having a direct impact on workplace efficiency and productivity. As <a href="https://www.itpro.com/technology/artificial-intelligence/workers-are-wasting-half-a-day-each-week-fixing-ai-workslop"><u><em>ITPro </em></u><u>reported in January</u></a>, analysis from Zapier found employees are working an extra four and a half hours each week cleaning up mistakes. </p><p>The consequences of not acting on low quality AI-generated outputs can be grave, the study noted. Respondents reported having had work rejected due to poor AI outputs, while others highlighted customer complaints and even security incidents. </p><h2 id="ai-is-adding-pressure">AI is adding pressure</h2><p>Workers are also feeling pressured, with half believing that their performance is now being compared – fairly or not – to AI-generated output. </p><p>Around a quarter are facing intense pressure to improve performance, improve quality, and be more efficient, just to keep pace in a machine-accelerated environment. </p><p>Another quarter said they use AI simply to meet workload demands, and 23% rely on the technology to keep up with colleagues.</p><p>"By understanding the nature of the work and the different value streams across your business, you can more accurately measure outcomes and impact rather than simply counting actions," said Riley. </p><p>"When AI is introduced thoughtfully, with the right guardrails and genuine support for the people using it, it can enhance rather than erode what makes work meaningful and impactful."</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Satya Nadella touts in-house models to cut spiralling AI costs ]]></title>
                                                                                                <dc:content><![CDATA[ <p><a href="https://www.itpro.com/technology/artificial-intelligence/satya-nadella-microsoft-ai-slop-2026">Microsoft CEO Satya Nadella </a>says the company plans to expand use of its MAI model range as the company looks to offer customers lower-cost AI options. </p><p>The in-house models, unveiled by the company in June, are designed for specific enterprise tasks and span a range of areas, including image and voice generation, audio transcription, and coding. </p><p>In a <a href="https://x.com/satyanadella/article/2080329851127669104" target="_blank"><u>blog post</u></a> on 23 July, Nadella outlined the tech giant’s new ‘Frontier Diffusion and Control’ strategy, which aims to help customers reduce reliance on costly frontier models and emphasize the importance of model choice for specific use-cases.</p><p>The move by Microsoft comes amidst growing concerns about spiralling AI costs over the last six months. With trends such as ‘tokenmaxxing’ and the shift toward consumption-based pricing, some companies have been left with hefty bills as AI use accelerates. </p><p>“In a world where software has real marginal cost for the first time, how do we ensure frontier benefits are diffused across the entire ecosystem?” Nadella wrote. </p><p>“The key is to optimize the cost-to-outcome frontier in real-world context,” he added. “In practical terms, that means using the right model for each task, and optimizing the context, skills, tools, and agent harness around it.”</p><h2 id="boxing-clever-with-ai-model-selection">Boxing clever with AI model selection</h2><p>Nadella’s contention here is that these in-house models can give customers “frontier capabilities” albeit in a cheaper, bespoke capacity. </p><p>Simply put, these are domain-specific models based for an enterprise’s individual needs, not a one-size-fits-all frontier model such as those offered by OpenAI or Anthropic. </p><p>These still form part of the broader “orchestration system” used alongside MAI, but all amounts to helping users box clever when choosing what model to use. </p><p>“These models have been built ground up with clean data lineage and optimized for learning transfer from generalist to specialized skills in enterprise RLEs,” he explained. </p><p>This will require enterprises to implement evaluation processes when considering which model to use in which context, he said. </p><p>“Product-specific evals and model independence give us the control and a direct hill to climb, and to keep refining until we reach the right quality-cost target,” Nadella wrote. </p><p>“We are now seeing MAI models outperform general-purpose frontier models in many use-cases while using a fraction of the tokens.”</p><p>Internal testing of MAI for its own products has so far delivered “promising early results”, according to Nadella. The company has piloted the use of these models in <a href="https://www.itpro.com/software/development/github-copilot-pricing-changes-usage-based-billing-explained">GitHub Copilot</a>, Outlook, and other <a href="https://www.itpro.com/desktop-software/19337/office-365-review">Microsoft 365</a> services. </p><p>The company plans to take the same approach with Copilot Chat, PowerPoint, and other services, he added. </p><p>Nadella’s comments align closely with recent research from Gartner on <a href="https://www.itpro.com/technology/artificial-intelligence/what-were-seeing-right-now-is-just-rapid-escalation-in-ai-token-spend-accenture-tells-staff-to-stop-using-ai-for-unnecessary-tasks-amid-surging-costs">rising AI costs</a>. </p><p>While focusing primarily on the use of AI in software engineering, <a href="https://www.itpro.com/software/development/surging-ai-costs-could-exceed-developer-salaries-by-2028-analysts-say-context-engineering-could-be-the-key-to-optimizing-token-consumption"><u>the consultancy told </u><u><em>ITPro</em></u></a><em> </em>that enterprises should establish a “use-case-driven decision” framework when using the technology for certain tasks. </p><p>This, Gartner noted, will help reduce costs by encouraging users to stop needlessly allocating agents to tasks they’re not designed for. </p><h2 id="microsoft-eyes-model-diversity">Microsoft eyes model diversity</h2><p>Nadella’s blog post marks the latest in a string of comments made by the Microsoft chief hinting at the company’s efforts to diversify model choice - and to stop pushing frontier models on customers. </p><p>In mid-July, he warned about <a href="https://www.itpro.com/technology/artificial-intelligence/a-company-should-be-able-to-use-a-model-without-giving-up-the-knowledge-that-makes-it-unique-microsoft-ceo-satya-nadella-says-enterprises-shouldnt-be-sharing-so-much-data-with-ai-providers"><u>the risks of relying too heavily on frontier AI labs</u></a>, describing what he called the “reverse information paradox”.</p><p>Nadella said enterprises are essentially “paying twice” for AI services, handing too much data to providers, and receiving little benefit in return. </p><p>These providers, meanwhile, gain valuable insights into customer products that could help them refine - or build - competing options.</p><p>"If learning flows in only one direction, economic value converges toward the owners of the learning infrastructure rather than the creators of the knowledge itself," he said. </p><p>"Therefore, it’s imperative that we distribute the learning infrastructure to every firm so that they can control their own learning loop.</p><p>As <em>ITPro </em>noted at the time, his comments were peculiar given the close ties the tech giant maintains with OpenAI and Anthropic, whose models are deeply entwined across its core product range. </p><p>Microsoft's long-standing relationship with OpenAI saw it take the lead as the go-to model for its software services, yet this changed last year after <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-has-a-new-ai-poster-child-in-anthropic">striking a deal with Anthropic</a> to give users greater choice. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/we-are-now-seeing-mai-models-outperform-general-purpose-frontier-models-microsoft-ceo-satya-nadella-touts-in-house-models-to-cut-spiralling-ai-costs-and-reduce-growing-reliance-on-frontier-labs</link>
                                                                            <description>
                            <![CDATA[ The Microsoft chief says pricey frontier models don't have to be used for every task, and its own in-house MAI models could be the key to reducing costs. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">ZakY85xsjj4Ady9Rjgw9TL</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/QojwEWhV4QWFHcEFxtEk2Z-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 24 Jul 2026 14:44:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:42 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/QojwEWhV4QWFHcEFxtEk2Z-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Microsoft CEO Satya Nadella pictured speaking on stage at the Microsoft AI Tour at TikTok Entertainment Centre on April 23, 2026 in Sydney, Australia.]]></media:description>                                                            <media:text><![CDATA[Microsoft CEO Satya Nadella pictured speaking on stage at the Microsoft AI Tour at TikTok Entertainment Centre on April 23, 2026 in Sydney, Australia.]]></media:text>
                                <media:title type="plain"><![CDATA[Microsoft CEO Satya Nadella pictured speaking on stage at the Microsoft AI Tour at TikTok Entertainment Centre on April 23, 2026 in Sydney, Australia.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/QojwEWhV4QWFHcEFxtEk2Z-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p><a href="https://www.itpro.com/technology/artificial-intelligence/satya-nadella-microsoft-ai-slop-2026">Microsoft CEO Satya Nadella </a>says the company plans to expand use of its MAI model range as the company looks to offer customers lower-cost AI options. </p><p>The in-house models, unveiled by the company in June, are designed for specific enterprise tasks and span a range of areas, including image and voice generation, audio transcription, and coding. </p><p>In a <a href="https://x.com/satyanadella/article/2080329851127669104" target="_blank"><u>blog post</u></a> on 23 July, Nadella outlined the tech giant’s new ‘Frontier Diffusion and Control’ strategy, which aims to help customers reduce reliance on costly frontier models and emphasize the importance of model choice for specific use-cases.</p><p>The move by Microsoft comes amidst growing concerns about spiralling AI costs over the last six months. With trends such as ‘tokenmaxxing’ and the shift toward consumption-based pricing, some companies have been left with hefty bills as AI use accelerates. </p><p>“In a world where software has real marginal cost for the first time, how do we ensure frontier benefits are diffused across the entire ecosystem?” Nadella wrote. </p><p>“The key is to optimize the cost-to-outcome frontier in real-world context,” he added. “In practical terms, that means using the right model for each task, and optimizing the context, skills, tools, and agent harness around it.”</p><h2 id="boxing-clever-with-ai-model-selection">Boxing clever with AI model selection</h2><p>Nadella’s contention here is that these in-house models can give customers “frontier capabilities” albeit in a cheaper, bespoke capacity. </p><p>Simply put, these are domain-specific models based for an enterprise’s individual needs, not a one-size-fits-all frontier model such as those offered by OpenAI or Anthropic. </p><p>These still form part of the broader “orchestration system” used alongside MAI, but all amounts to helping users box clever when choosing what model to use. </p><p>“These models have been built ground up with clean data lineage and optimized for learning transfer from generalist to specialized skills in enterprise RLEs,” he explained. </p><p>This will require enterprises to implement evaluation processes when considering which model to use in which context, he said. </p><p>“Product-specific evals and model independence give us the control and a direct hill to climb, and to keep refining until we reach the right quality-cost target,” Nadella wrote. </p><p>“We are now seeing MAI models outperform general-purpose frontier models in many use-cases while using a fraction of the tokens.”</p><p>Internal testing of MAI for its own products has so far delivered “promising early results”, according to Nadella. The company has piloted the use of these models in <a href="https://www.itpro.com/software/development/github-copilot-pricing-changes-usage-based-billing-explained">GitHub Copilot</a>, Outlook, and other <a href="https://www.itpro.com/desktop-software/19337/office-365-review">Microsoft 365</a> services. </p><p>The company plans to take the same approach with Copilot Chat, PowerPoint, and other services, he added. </p><p>Nadella’s comments align closely with recent research from Gartner on <a href="https://www.itpro.com/technology/artificial-intelligence/what-were-seeing-right-now-is-just-rapid-escalation-in-ai-token-spend-accenture-tells-staff-to-stop-using-ai-for-unnecessary-tasks-amid-surging-costs">rising AI costs</a>. </p><p>While focusing primarily on the use of AI in software engineering, <a href="https://www.itpro.com/software/development/surging-ai-costs-could-exceed-developer-salaries-by-2028-analysts-say-context-engineering-could-be-the-key-to-optimizing-token-consumption"><u>the consultancy told </u><u><em>ITPro</em></u></a><em> </em>that enterprises should establish a “use-case-driven decision” framework when using the technology for certain tasks. </p><p>This, Gartner noted, will help reduce costs by encouraging users to stop needlessly allocating agents to tasks they’re not designed for. </p><h2 id="microsoft-eyes-model-diversity">Microsoft eyes model diversity</h2><p>Nadella’s blog post marks the latest in a string of comments made by the Microsoft chief hinting at the company’s efforts to diversify model choice - and to stop pushing frontier models on customers. </p><p>In mid-July, he warned about <a href="https://www.itpro.com/technology/artificial-intelligence/a-company-should-be-able-to-use-a-model-without-giving-up-the-knowledge-that-makes-it-unique-microsoft-ceo-satya-nadella-says-enterprises-shouldnt-be-sharing-so-much-data-with-ai-providers"><u>the risks of relying too heavily on frontier AI labs</u></a>, describing what he called the “reverse information paradox”.</p><p>Nadella said enterprises are essentially “paying twice” for AI services, handing too much data to providers, and receiving little benefit in return. </p><p>These providers, meanwhile, gain valuable insights into customer products that could help them refine - or build - competing options.</p><p>"If learning flows in only one direction, economic value converges toward the owners of the learning infrastructure rather than the creators of the knowledge itself," he said. </p><p>"Therefore, it’s imperative that we distribute the learning infrastructure to every firm so that they can control their own learning loop.</p><p>As <em>ITPro </em>noted at the time, his comments were peculiar given the close ties the tech giant maintains with OpenAI and Anthropic, whose models are deeply entwined across its core product range. </p><p>Microsoft's long-standing relationship with OpenAI saw it take the lead as the go-to model for its software services, yet this changed last year after <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-has-a-new-ai-poster-child-in-anthropic">striking a deal with Anthropic</a> to give users greater choice. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Nvidia CEO Jensen Huang hails 'excellent' Chinese AI models ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Nvidia CEO Jensen Huang has hailed the capabilities of Chinese open source AI models amidst growing interest in low-cost options for enterprises. </p><p>In an interview with <a href="https://www.axios.com/2026/07/22/nvidia-jensen-huang-china-open-source-ai" target="_blank"><u><em>Axios </em></u></a>this week, the Nvidia chief said competition between open and closed-source models is positive for the broader global AI industry. </p><p>“The world needs open models,” he told Axios. “These Chinese models are excellent. <a href="https://www.itpro.com/technology/artificial-intelligence/three-open-source-large-language-models-you-can-use-today">Open source models</a> that are excellent should be used”. </p><p>Huang’s comments come amidst growing interest in Chinese open source AI models in recent weeks. The Kimi K3 model, developed by Moonshot AI, temporarily paused subscriptions due to skyrocketing demand. </p><p>Surging interest in the Kimi K3 model bears similarities to that of DeepSeek, which <a href="https://www.itpro.com/technology/artificial-intelligence/the-deepseek-bombshell-has-been-a-wakeup-call-for-us-tech-giants">rocked the US AI industry </a>when it launched in early 2025. </p><p>A key advantage of these models lies in both their performance and cost efficiency, the latter of which has become a key focus for enterprises amidst growing concerns over spiralling AI costs in recent months. </p><p>As <a href="https://www.itpro.com/software/open-source/open-source-ai-performance-cost-savings-proprietary-models-linux-foundation"><u><em>ITPro </em></u><u>reported in November 2025</u></a>, open source AI models often perform on-par with closed source options, and could potentially save enterprises millions in costs each year.  </p><p>Open source models "routinely achieve 90% or more” of the performance of closed counterparts, the study noted. They also benefit from “significantly lower prices”, with operational costs plunging up to 84% in some cases. </p><p>In a <a href="https://go.redirectingat.com/?id=92X1583683&amp;xcust=itpro_gb_1425528928358339126&amp;xs=1&amp;url=https%3A%2F%2Fwww.nature.com%2Farticles%2Fs41586-025-09422-z&amp;sref=https%3A%2F%2Fwww.itpro.com" target="_blank"><u>research paper</u></a> published by DeepSeek last year, the company said it <a href="https://www.itpro.com/technology/artificial-intelligence/deepseeks-r1-model-training-costs-pour-cold-water-on-big-techs-massive-ai-spending"><u>spent a paltry amount training its flagship R1 model</u></a> compared to US rivals who’ve spent billions on training purposes. </p><p>Anthropic CEO Dario Amodei suggested <a href="https://www.itpro.com/technology/artificial-intelligence/dollar100-billion-to-build-an-ai-model-anthropic-ceo-dario-amodei-predicts-soaring-ai-training-costs-but-models-will-become-far-more-powerful"><u>building a new AI model could cost over $100 billion</u></a> in 2024. </p><h2 id="security-concerns-persist">Security concerns persist</h2><p>While the popularity of Chinese AI models continues rising, concerns over security have been flagged repeatedly. </p><p>The concerns surrounding Chinese AI models typically focus on areas such as data privacy and national security, with some critics suggesting these tools could be used as a ‘backdoor’ for Chinese intelligence services. </p><p><a href="https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi" target="_blank"><u><em>Axios </em></u><u>previously reported</u></a> that the White House could consider imposing restrictions or tight conditions on US firms working with these models. </p><p>DeepSeek was <a href="https://www.itpro.com/technology/artificial-intelligence/deepseek-r1-model-jailbreak-security-flaws">subject to the same scrutiny</a> during its rise to prominence across 2025, for example. In August last year, <a href="https://www.itpro.com/security/using-deepseek-at-work-security-risks"><u>cybersecurity experts told </u><u><em>ITPro</em></u></a><em> </em>that using DeepSeek equated to “printing out and handing over your confidential information”. </p><p>Huang told Axios that these concerns are a “misconception” and can be used by enterprises safely. On the topic of potential White House restrictions, Huang said this could prove counterproductive and that American firms should consider a breadth of options. </p><p>“You can download the models. You could fine-tune it, you can enhance it, you can guardrail it as you desire,” he said. </p><p>Huang has a vested interest on this front, however, and noted that increased usage will ultimately benefit Nvidia. </p><p>“If there’s great AI, even if it’s open, wherever it comes from, there will be more use,” he told <em>Axios</em>. “Whenever there’s more use, you’ll have to sell a lot more Nvidia computers. We’ll have to build more data centers. We’ll have more services. The technology will diffuse into more industries”. </p><p>Bill Conner, president and CEO Jitterbit, said the Kimi K3 launch marks a “clear turning point in accelerating global adoption of Chinese LLMs” but still urged caution with regard to security. </p><p>“Enterprises may underestimate the security, accountability, and governance risks associated with adopting new Chinese LLMs,” he said. </p><p>“Moonshot AI operates as a cloud service with data stored in China, introducing potential risks for organizations whose data may be shared unknowingly.”</p><p>Conner said adoption of Chinese AI models should be treated as a “strategic risk decision, not merely a cost-saving opportunity”. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/software/open-source/these-chinese-models-are-excellent-nvidia-ceo-jensen-huang-hails-powerful-new-chinese-ai-models-like-kimi-k3-and-says-dont-be-put-off-by-security-misconceptions</link>
                                                                            <description>
                            <![CDATA[ As powerful new AI models like Kimi K3 hit the market, Huang says competition will be a positive for the global industry ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">TPXwaJQQSfDmtup4qWjdaF</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/Nx4Vkmnsn4FBGv225qYypT-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 23 Jul 2026 11:57:23 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:43 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Open Source]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                    <category><![CDATA[Workplace Technology]]></category>
                                                    <category><![CDATA[Software]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/Nx4Vkmnsn4FBGv225qYypT-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Nvidia CEO Jensen Huang pictured in his signature leather jacket giving a thumbs up during a photo session following a joint press conference with representatives of Fujitsu.]]></media:description>                                                            <media:text><![CDATA[Nvidia CEO Jensen Huang pictured in his signature leather jacket giving a thumbs up during a photo session following a joint press conference with representatives of Fujitsu.]]></media:text>
                                <media:title type="plain"><![CDATA[Nvidia CEO Jensen Huang pictured in his signature leather jacket giving a thumbs up during a photo session following a joint press conference with representatives of Fujitsu.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/Nx4Vkmnsn4FBGv225qYypT-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Nvidia CEO Jensen Huang has hailed the capabilities of Chinese open source AI models amidst growing interest in low-cost options for enterprises. </p><p>In an interview with <a href="https://www.axios.com/2026/07/22/nvidia-jensen-huang-china-open-source-ai" target="_blank"><u><em>Axios </em></u></a>this week, the Nvidia chief said competition between open and closed-source models is positive for the broader global AI industry. </p><p>“The world needs open models,” he told Axios. “These Chinese models are excellent. <a href="https://www.itpro.com/technology/artificial-intelligence/three-open-source-large-language-models-you-can-use-today">Open source models</a> that are excellent should be used”. </p><p>Huang’s comments come amidst growing interest in Chinese open source AI models in recent weeks. The Kimi K3 model, developed by Moonshot AI, temporarily paused subscriptions due to skyrocketing demand. </p><p>Surging interest in the Kimi K3 model bears similarities to that of DeepSeek, which <a href="https://www.itpro.com/technology/artificial-intelligence/the-deepseek-bombshell-has-been-a-wakeup-call-for-us-tech-giants">rocked the US AI industry </a>when it launched in early 2025. </p><p>A key advantage of these models lies in both their performance and cost efficiency, the latter of which has become a key focus for enterprises amidst growing concerns over spiralling AI costs in recent months. </p><p>As <a href="https://www.itpro.com/software/open-source/open-source-ai-performance-cost-savings-proprietary-models-linux-foundation"><u><em>ITPro </em></u><u>reported in November 2025</u></a>, open source AI models often perform on-par with closed source options, and could potentially save enterprises millions in costs each year.  </p><p>Open source models "routinely achieve 90% or more” of the performance of closed counterparts, the study noted. They also benefit from “significantly lower prices”, with operational costs plunging up to 84% in some cases. </p><p>In a <a href="https://go.redirectingat.com/?id=92X1583683&amp;xcust=itpro_gb_1425528928358339126&amp;xs=1&amp;url=https%3A%2F%2Fwww.nature.com%2Farticles%2Fs41586-025-09422-z&amp;sref=https%3A%2F%2Fwww.itpro.com" target="_blank"><u>research paper</u></a> published by DeepSeek last year, the company said it <a href="https://www.itpro.com/technology/artificial-intelligence/deepseeks-r1-model-training-costs-pour-cold-water-on-big-techs-massive-ai-spending"><u>spent a paltry amount training its flagship R1 model</u></a> compared to US rivals who’ve spent billions on training purposes. </p><p>Anthropic CEO Dario Amodei suggested <a href="https://www.itpro.com/technology/artificial-intelligence/dollar100-billion-to-build-an-ai-model-anthropic-ceo-dario-amodei-predicts-soaring-ai-training-costs-but-models-will-become-far-more-powerful"><u>building a new AI model could cost over $100 billion</u></a> in 2024. </p><h2 id="security-concerns-persist">Security concerns persist</h2><p>While the popularity of Chinese AI models continues rising, concerns over security have been flagged repeatedly. </p><p>The concerns surrounding Chinese AI models typically focus on areas such as data privacy and national security, with some critics suggesting these tools could be used as a ‘backdoor’ for Chinese intelligence services. </p><p><a href="https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi" target="_blank"><u><em>Axios </em></u><u>previously reported</u></a> that the White House could consider imposing restrictions or tight conditions on US firms working with these models. </p><p>DeepSeek was <a href="https://www.itpro.com/technology/artificial-intelligence/deepseek-r1-model-jailbreak-security-flaws">subject to the same scrutiny</a> during its rise to prominence across 2025, for example. In August last year, <a href="https://www.itpro.com/security/using-deepseek-at-work-security-risks"><u>cybersecurity experts told </u><u><em>ITPro</em></u></a><em> </em>that using DeepSeek equated to “printing out and handing over your confidential information”. </p><p>Huang told Axios that these concerns are a “misconception” and can be used by enterprises safely. On the topic of potential White House restrictions, Huang said this could prove counterproductive and that American firms should consider a breadth of options. </p><p>“You can download the models. You could fine-tune it, you can enhance it, you can guardrail it as you desire,” he said. </p><p>Huang has a vested interest on this front, however, and noted that increased usage will ultimately benefit Nvidia. </p><p>“If there’s great AI, even if it’s open, wherever it comes from, there will be more use,” he told <em>Axios</em>. “Whenever there’s more use, you’ll have to sell a lot more Nvidia computers. We’ll have to build more data centers. We’ll have more services. The technology will diffuse into more industries”. </p><p>Bill Conner, president and CEO Jitterbit, said the Kimi K3 launch marks a “clear turning point in accelerating global adoption of Chinese LLMs” but still urged caution with regard to security. </p><p>“Enterprises may underestimate the security, accountability, and governance risks associated with adopting new Chinese LLMs,” he said. </p><p>“Moonshot AI operates as a cloud service with data stored in China, introducing potential risks for organizations whose data may be shared unknowingly.”</p><p>Conner said adoption of Chinese AI models should be treated as a “strategic risk decision, not merely a cost-saving opportunity”. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ What users can expect with Claude Sonnet 5 ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Anthropic has launched Claude Sonnet 5 in a move the company said offers users more powerful agentic capabilities. </p><p>The new option is the “most agentic Sonnet model yet”, according to Anthropic, capable of building workflow plans for users, accessing browsers and terminals, and operating autonomously for longer periods.</p><p>Cost and performance are key talking points here for Anthropic, with the company noting that Sonnet 5 “narrows the gap” with its flagship Opus range in terms of efficiency. </p><p>“Claude Sonnet 3.5, 3.6, and 3.7 were the first models that showed impressive skills in coding and tool use. More recently, though, the clearest gains in agentic capabilities have been in our Opus-class models,” the company said in a <a href="https://www.anthropic.com/news/claude-sonnet-5" target="_blank"><u>blog post</u></a>. </p><p>“Its performance is close to that of Opus 4.8, but at lower prices,” Anthropic added. “It’s a substantial improvement over its predecessor, <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-promises-opus-level-reasoning-claude-sonnet-4-6-model-at-lower-cost">Sonnet 4.6</a>, on important aspects of agentic performance like reasoning, tool use, <a href="https://www.itpro.com/technology/artificial-intelligence/vibe-coding-security-risks-how-to-mitigate">coding</a>, and knowledge work.”</p><p>SWE-bench Pro results show that show that Sonnet 5 recorded a 63.2% score on agentic coding capabilities. That marks a solid increase compared to Sonnet 4.6, which came in at 58.1%. </p><p>The model still lags behind Opus 4.8 (69.2%), but clear gains are being made on this front. </p><h2 id="anthropic-eyes-cost-improvements-with-claude-sonnet-5">Anthropic eyes cost improvements with Claude Sonnet 5</h2><p>With concerns mounting over token consumption, Anthropic is keen to emphasize that Sonnet is a far more cost-efficient model. </p><p>Performance of the model at “different effort levels” – or the intensity and timeframe of tasks – on BrowseComp shows a “strict improvement” over Sonnet 4.6. </p><p>Compared to Opus 4.8, meanwhile, Sonnet covers a “much wider range of cost-performance options”. </p><p>“It provides substantially improved cost efficiency at medium effort,” the company explained. “Its higher-effort performance can match Opus 4.8 on some tasks. </p><p>Users will also be able to adjust effort levels to help them “find the right balance of cost and performance” with both Sonnet 5 and Opus 4.8. </p><h2 id="sonnet-5-safety-improvements">Sonnet 5 safety improvements</h2><p>On the safety front, Anthropic noted there has been an “overall improvement” with Sonnet 5, which is far more likely to refuse “malicious requests” than 4.8 and is capable of resisting hijack attempts, such as <a href="https://www.itpro.com/security/ncsc-issues-urgent-warning-over-growing-ai-prompt-injection-risks-heres-what-you-need-to-know">prompt injection</a>. </p><p>“The model shows lower rates of hallucination and sycophancy than Sonnet 4.6,” the company noted. “On our automated behavioral audit, which tests a wide range of misaligned behaviors such as cooperation with misuse and deception, Sonnet 5 scored lower (that is, safer) overall.”</p><p>It’s worth noting that Anthropic admits Sonnet 5 displays higher rates of “misaligned behavior” compared to Opus 4.8 and Claude Mythos, urging caution by users. </p><p>Safety tests for Sonnet 5 come at a critical time for Anthropic, with the company having launched <a href="https://www.itpro.com/technology/artificial-intelligence/project-glasswing-anthropic-announces-big-tech-consortium-to-test-claude-mythos-ai-model-that-could-reshape-cybersecurity">Claude Mythos</a> earlier this year as part of a gated preview with select industry partners. </p><h2 id="how-to-access-sonnet-5">How to access Sonnet 5</h2><p>Claude Sonnet 5 is available across all plans from today, according to Anthropic, including Max, Team, and Enterprise users. It will also act as the default model for Free and Pro plans. </p><p>For <a href="https://www.itpro.com/software/development/anthropic-claude-code-usage-limits-increase-spacex-compute-deal">Claude Code</a> and Claude Platform users, Sonnet will come with an “introductory” pricing of $2 per million input tokens, and $10 per million output tokens until 31 August. </p><p>Thereafter, Anthropic said prices will rise to $3 and $15 across input and output tokens respectively. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/anthropic-touts-new-claude-sonnet-5-model-range-offering-performance-close-to-that-of-opus-4-8-but-at-lower-prices-heres-what-users-can-expect</link>
                                                                            <description>
                            <![CDATA[ Claude Sonnet 5 comes with intuitive agentic capabilities, performance boosts, and cost-efficient ‘effort levels’ ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">AnRXw6w2eziVMdMG9vo6Jf</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/vHLUVkTVoTfgLui95XScvD-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 01 Jul 2026 15:55:30 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:48 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/vHLUVkTVoTfgLui95XScvD-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Logo and branding of Anthropic&#039;s Claude AI tool pictured on a smartphone screen, with branding blending into background.]]></media:description>                                                            <media:text><![CDATA[Logo and branding of Anthropic&#039;s Claude AI tool pictured on a smartphone screen, with branding blending into background.]]></media:text>
                                <media:title type="plain"><![CDATA[Logo and branding of Anthropic&#039;s Claude AI tool pictured on a smartphone screen, with branding blending into background.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/vHLUVkTVoTfgLui95XScvD-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Anthropic has launched Claude Sonnet 5 in a move the company said offers users more powerful agentic capabilities. </p><p>The new option is the “most agentic Sonnet model yet”, according to Anthropic, capable of building workflow plans for users, accessing browsers and terminals, and operating autonomously for longer periods.</p><p>Cost and performance are key talking points here for Anthropic, with the company noting that Sonnet 5 “narrows the gap” with its flagship Opus range in terms of efficiency. </p><p>“Claude Sonnet 3.5, 3.6, and 3.7 were the first models that showed impressive skills in coding and tool use. More recently, though, the clearest gains in agentic capabilities have been in our Opus-class models,” the company said in a <a href="https://www.anthropic.com/news/claude-sonnet-5" target="_blank"><u>blog post</u></a>. </p><p>“Its performance is close to that of Opus 4.8, but at lower prices,” Anthropic added. “It’s a substantial improvement over its predecessor, <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-promises-opus-level-reasoning-claude-sonnet-4-6-model-at-lower-cost">Sonnet 4.6</a>, on important aspects of agentic performance like reasoning, tool use, <a href="https://www.itpro.com/technology/artificial-intelligence/vibe-coding-security-risks-how-to-mitigate">coding</a>, and knowledge work.”</p><p>SWE-bench Pro results show that show that Sonnet 5 recorded a 63.2% score on agentic coding capabilities. That marks a solid increase compared to Sonnet 4.6, which came in at 58.1%. </p><p>The model still lags behind Opus 4.8 (69.2%), but clear gains are being made on this front. </p><h2 id="anthropic-eyes-cost-improvements-with-claude-sonnet-5">Anthropic eyes cost improvements with Claude Sonnet 5</h2><p>With concerns mounting over token consumption, Anthropic is keen to emphasize that Sonnet is a far more cost-efficient model. </p><p>Performance of the model at “different effort levels” – or the intensity and timeframe of tasks – on BrowseComp shows a “strict improvement” over Sonnet 4.6. </p><p>Compared to Opus 4.8, meanwhile, Sonnet covers a “much wider range of cost-performance options”. </p><p>“It provides substantially improved cost efficiency at medium effort,” the company explained. “Its higher-effort performance can match Opus 4.8 on some tasks. </p><p>Users will also be able to adjust effort levels to help them “find the right balance of cost and performance” with both Sonnet 5 and Opus 4.8. </p><h2 id="sonnet-5-safety-improvements">Sonnet 5 safety improvements</h2><p>On the safety front, Anthropic noted there has been an “overall improvement” with Sonnet 5, which is far more likely to refuse “malicious requests” than 4.8 and is capable of resisting hijack attempts, such as <a href="https://www.itpro.com/security/ncsc-issues-urgent-warning-over-growing-ai-prompt-injection-risks-heres-what-you-need-to-know">prompt injection</a>. </p><p>“The model shows lower rates of hallucination and sycophancy than Sonnet 4.6,” the company noted. “On our automated behavioral audit, which tests a wide range of misaligned behaviors such as cooperation with misuse and deception, Sonnet 5 scored lower (that is, safer) overall.”</p><p>It’s worth noting that Anthropic admits Sonnet 5 displays higher rates of “misaligned behavior” compared to Opus 4.8 and Claude Mythos, urging caution by users. </p><p>Safety tests for Sonnet 5 come at a critical time for Anthropic, with the company having launched <a href="https://www.itpro.com/technology/artificial-intelligence/project-glasswing-anthropic-announces-big-tech-consortium-to-test-claude-mythos-ai-model-that-could-reshape-cybersecurity">Claude Mythos</a> earlier this year as part of a gated preview with select industry partners. </p><h2 id="how-to-access-sonnet-5">How to access Sonnet 5</h2><p>Claude Sonnet 5 is available across all plans from today, according to Anthropic, including Max, Team, and Enterprise users. It will also act as the default model for Free and Pro plans. </p><p>For <a href="https://www.itpro.com/software/development/anthropic-claude-code-usage-limits-increase-spacex-compute-deal">Claude Code</a> and Claude Platform users, Sonnet will come with an “introductory” pricing of $2 per million input tokens, and $10 per million output tokens until 31 August. </p><p>Thereafter, Anthropic said prices will rise to $3 and $15 across input and output tokens respectively. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ OpenAI plots ChatGPT ‘super app’ overhaul ahead of public listing ]]></title>
                                                                                                <dc:content><![CDATA[ <p>OpenAI is reportedly planning a major redesign of <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369965/what-is-chatgpt-and-what-does-it-mean-for-businesses">ChatGPT</a>, with the company set to sharpen its focus on agents and coding capabilities. </p><p>ChatGPT's launch in 2022 kickstarted the AI boom, and it remains one of the most popular <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> applications. But reports suggest OpenAI is looking to post stronger revenue, in particular ahead of a potential initial public offering (IPO).</p><p>The plans, as reported by the <a href="https://www.ft.com/content/ca0f5f5e-fb9a-41a0-a2a9-0127e15b7db9?syn-25a6b1a6=1" target="_blank"><u><em>Financial Times</em></u><u>,</u></a> include the creation of a "super app" that includes coding tools and an AI agent, replacing the traditional chatbot-focused format of the highly popular tool. </p><p>The move is designed to help drive revenue, sources told the publication, a front on which OpenAI has been pushing heavily over the last 18 months. </p><h2 id="chat-is-dead">"Chat is dead"</h2><p>OpenAI's plans could see the company shift its focus away from simple chat, as the <em>FT </em>reports executives at the company view it as merely a gateway to signing up users to paid-for products. </p><p><a href="https://www.reuters.com/technology/chatgpt-app-hits-1-billion-monthly-active-users-record-time-data-shows-2026-06-02/" target="_blank"><u>ChatGPT has a billion monthly users</u></a> globally, according to Sensor Tower data, though most of those are free rather than paid users, with roughly two million business customers globally. </p><p>The newspaper quoted one employee as saying "chat is dead" – hinting at the future direction of the platform. </p><p>Instead of chat, the aim is to convince users to step up to using AI agents to complete tasks, rather than merely using the chatbot for information. The so-called "superapp" will also include <a href="https://www.itpro.com/business/business-strategy/openais-codex-developer-agent-just-got-a-big-update">OpenAI's Codex coding product</a>, which allows users to create their own custom agents. </p><p>Changes to ChatGPT are expected to begin rolling out in the coming weeks, according to the report. However, rather than a total overhaul, these will surface as small changes to the app and website that will push users towards other tools for coding, image generation, and even third-party apps. </p><p>The aim, sources told the <em>FT</em>, is to build a personal agent that can help users across work and their personal life, and be accessible everywhere – even in the car. </p><h2 id="openai-s-big-revenue-push">OpenAI’s big revenue push</h2><p>At the end of last year, OpenAI was rumoured to be <a href="https://www.itpro.com/business/business-strategy/openai-1-trillion-infrastructure-investment-oracle-broadcom-amd"><u>considering everything</u></a> from advertising to personal assistants as it sought to find ways to pay for more than $1 trillion in spending it had lined up at the time. </p><p>The company has also targeted corporate customers via a <a href="https://www.itpro.com/business/business-strategy/openai-ramps-up-enterprise-ai-push-with-new-consultancy-launch"><u>new consultancy designed</u></a> to make it easier to roll out its tools in enterprises. At the same time, OpenAI shuttered the video generation app Sora, as part of efforts to focus on the new app. </p><p>Earlier this year, OpenAI <a href="https://www.itpro.com/business/business-strategy/openai-cfo-teases-new-money-making-plans-as-ai-giant-targets-practical-adoption-gains-in-2026-and-beyond"><u>said the company's revenue</u></a> topped $20 billion in 2025, up from $6bn the year before. CFO Sarah Friar said at the time that was "never-before-seen growth at such scale." </p><p>But that comes with high costs as well as a string of multi-billion dollar infrastructure deals with <a href="https://www.itpro.com/cloud/cloud-computing/openai-oracle-cloud-deal-stargate">industry counterparts such as Oracle</a>. </p><p>Competition is also heating up for OpenAI, particularly from Anthropic. The key rival recently confidentially filed to make a public debut, and has been the subject of intense corporate attention <a href="https://www.itpro.com/technology/artificial-intelligence/why-anthropic-sent-software-stocks-into-freefall"><u>targeting key services and software</u></a> in recent months. </p><p>Jenny Xiao, partner at Leonis Capital and former researcher at OpenAI, told the <em>FT </em>that OpenAI started by trying to gain as many users as possible, while Anthropic targeted revenue. </p><p>"Now the two are converging, because both of them are trying to aim for an IPO and investors care more about money than dreams," she said. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/openai-chatgpt-superapp-overhaul-public-listing</link>
                                                                            <description>
                            <![CDATA[ The company looks set to spruce up ChatGPT with a particular focus on agents to drive subscriptions ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">23dDVMi2JPUM3EaFkTGjfA</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/2pdMPJCfCPRpfB9M8widPN-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 10 Jun 2026 08:15:38 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:55 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/2pdMPJCfCPRpfB9M8widPN-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[OpenAI chief executive Sam Altman pictured speaking on stage during a panel discussion at the BlackRock Infrastructure Summit in Washington DC, United States. ]]></media:description>                                                            <media:text><![CDATA[OpenAI chief executive Sam Altman pictured speaking on stage during a panel discussion at the BlackRock Infrastructure Summit in Washington DC, United States. ]]></media:text>
                                <media:title type="plain"><![CDATA[OpenAI chief executive Sam Altman pictured speaking on stage during a panel discussion at the BlackRock Infrastructure Summit in Washington DC, United States. ]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/2pdMPJCfCPRpfB9M8widPN-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>OpenAI is reportedly planning a major redesign of <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369965/what-is-chatgpt-and-what-does-it-mean-for-businesses">ChatGPT</a>, with the company set to sharpen its focus on agents and coding capabilities. </p><p>ChatGPT's launch in 2022 kickstarted the AI boom, and it remains one of the most popular <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> applications. But reports suggest OpenAI is looking to post stronger revenue, in particular ahead of a potential initial public offering (IPO).</p><p>The plans, as reported by the <a href="https://www.ft.com/content/ca0f5f5e-fb9a-41a0-a2a9-0127e15b7db9?syn-25a6b1a6=1" target="_blank"><u><em>Financial Times</em></u><u>,</u></a> include the creation of a "super app" that includes coding tools and an AI agent, replacing the traditional chatbot-focused format of the highly popular tool. </p><p>The move is designed to help drive revenue, sources told the publication, a front on which OpenAI has been pushing heavily over the last 18 months. </p><h2 id="chat-is-dead">"Chat is dead"</h2><p>OpenAI's plans could see the company shift its focus away from simple chat, as the <em>FT </em>reports executives at the company view it as merely a gateway to signing up users to paid-for products. </p><p><a href="https://www.reuters.com/technology/chatgpt-app-hits-1-billion-monthly-active-users-record-time-data-shows-2026-06-02/" target="_blank"><u>ChatGPT has a billion monthly users</u></a> globally, according to Sensor Tower data, though most of those are free rather than paid users, with roughly two million business customers globally. </p><p>The newspaper quoted one employee as saying "chat is dead" – hinting at the future direction of the platform. </p><p>Instead of chat, the aim is to convince users to step up to using AI agents to complete tasks, rather than merely using the chatbot for information. The so-called "superapp" will also include <a href="https://www.itpro.com/business/business-strategy/openais-codex-developer-agent-just-got-a-big-update">OpenAI's Codex coding product</a>, which allows users to create their own custom agents. </p><p>Changes to ChatGPT are expected to begin rolling out in the coming weeks, according to the report. However, rather than a total overhaul, these will surface as small changes to the app and website that will push users towards other tools for coding, image generation, and even third-party apps. </p><p>The aim, sources told the <em>FT</em>, is to build a personal agent that can help users across work and their personal life, and be accessible everywhere – even in the car. </p><h2 id="openai-s-big-revenue-push">OpenAI’s big revenue push</h2><p>At the end of last year, OpenAI was rumoured to be <a href="https://www.itpro.com/business/business-strategy/openai-1-trillion-infrastructure-investment-oracle-broadcom-amd"><u>considering everything</u></a> from advertising to personal assistants as it sought to find ways to pay for more than $1 trillion in spending it had lined up at the time. </p><p>The company has also targeted corporate customers via a <a href="https://www.itpro.com/business/business-strategy/openai-ramps-up-enterprise-ai-push-with-new-consultancy-launch"><u>new consultancy designed</u></a> to make it easier to roll out its tools in enterprises. At the same time, OpenAI shuttered the video generation app Sora, as part of efforts to focus on the new app. </p><p>Earlier this year, OpenAI <a href="https://www.itpro.com/business/business-strategy/openai-cfo-teases-new-money-making-plans-as-ai-giant-targets-practical-adoption-gains-in-2026-and-beyond"><u>said the company's revenue</u></a> topped $20 billion in 2025, up from $6bn the year before. CFO Sarah Friar said at the time that was "never-before-seen growth at such scale." </p><p>But that comes with high costs as well as a string of multi-billion dollar infrastructure deals with <a href="https://www.itpro.com/cloud/cloud-computing/openai-oracle-cloud-deal-stargate">industry counterparts such as Oracle</a>. </p><p>Competition is also heating up for OpenAI, particularly from Anthropic. The key rival recently confidentially filed to make a public debut, and has been the subject of intense corporate attention <a href="https://www.itpro.com/technology/artificial-intelligence/why-anthropic-sent-software-stocks-into-freefall"><u>targeting key services and software</u></a> in recent months. </p><p>Jenny Xiao, partner at Leonis Capital and former researcher at OpenAI, told the <em>FT </em>that OpenAI started by trying to gain as many users as possible, while Anthropic targeted revenue. </p><p>"Now the two are converging, because both of them are trying to aim for an IPO and investors care more about money than dreams," she said. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Anthropic just launched Claude Fable 5, its first Mythos-class AI model ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Anthropic has unveiled a new Mythos-class AI model designed specifically for public use. </p><p>The new model, dubbed Fable 5, boasts significant performance improvements on previous models, with the company noting that it “exceeds every model” previously made available to the public.</p><p>Notably, Fable 5 has been tweaked to include new safeguards that “block responses in specific high-risk areas” to prevent misuse – a key concern with the <a href="https://www.itpro.com/technology/artificial-intelligence/project-glasswing-anthropic-announces-big-tech-consortium-to-test-claude-mythos-ai-model-that-could-reshape-cybersecurity">launch of Claude Mythos</a>. </p><p>Claude Mythos was rolled-out as part of a gated release in April this year, with industry partners such as Microsoft, Amazon, and Cisco selected to test the model’s capabilities behind closed doors. </p><p>With the launch of Fable 5, Anthropic revealed that the model will “fall back” to Claude Opus 4.8 in the event that potential high-risk uses are detected. </p><p>Alongside the launch of Fable 5, Anthropic also revealed it will release Claude Mythos 5, again in a restricted capacity. </p><p>“Beginning today, all users who currently have access to Claude Mythos Preview (for example, our cybersecurity partners in <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-lets-glasswing-partners-publicly-share-mythos-flaws">Project Glasswing</a>) will be able to upgrade to Claude Mythos 5,” the company said in an announcement. </p><p>The company added that it plans to expand access in due course through a “more systematic trusted-access program”. </p><h2 id="performance-boosts-with-claude-fable-5">Performance boosts with Claude Fable 5</h2><p>According to Anthropic, Fable 5 shows “exceptional performance” in software engineering-related tasks, as well as knowledge work and scientific research. </p><p>Testing of the model by analytics company Hex revealed the model was the first to record a score of over 90% on its core analytics benchmark, which consists of running the model through “complex, long-running analytical tasks”. </p><p>Indeed, Hex noted that Fable recorded a 10-point improvement over Opus. </p><p>Similar testing by agentic workspace platform, Genspark, also showed significant performance boosts compared to other models, particularly in areas such as coding. </p><p>“Fable came out #1 on our evals, winning head-to-head against every model we tested,” the company said. “It was significantly stronger on the hardest tasks in the set — UI design and game coding."</p><h2 id="robust-guardrails-for-fable-5">Robust guardrails for Fable 5</h2><p>With Fable 5’s new safeguards, the model will be restricted from performing tasks in a range of areas, including cybersecurity and biology, and will revert back to Opus 4.8 to provide answers to queries on these topics. </p><p>“Releasing a model this capable comes with risks,” Anthropic said in a statement. “Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage.”</p><p>Anthropic said it “extensively red-teamed” classifiers to test the model’s defenses against potential jailbreaks as part of a bug bounty. </p><p>“Internally, we ran an external bug bounty that produced no universal jailbreaks in over 1,000 hours of testing,” the company said. “We then worked with external red-teaming orgs which also failed to find universal jailbreaks.”</p><h2 id="fable-5-pricing-and-access">Fable 5 pricing and access</h2><p>According to Anthropic, pricing for both models will stand at $10 per million input tokens, with $50 per million output tokens. </p><p>Token consumption costs have become a hot topic in the tech industry over the last two months with the rise of ‘tokenmaxxing’. </p><p>Uber, for example, revealed it <a href="https://www.itpro.com/technology/artificial-intelligence/ubers-eye-watering-ai-bill-shows-enterprises-are-still-measuring-ai-success-through-consumption-rather-than-outcomes-and-its-warping-our-perception-of-roi-and-productivity">blew through its entire annual AI budget in just four months</a> amidst skyrocketing use by staff. </p><p>Developers will be able to use Fable 5 through the Claude API while Mythos 5 will only be accessible to Project Glasswing participants. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/anthropic-just-launched-claude-fable-5-its-first-mythos-class-ai-model-but-it-has-new-safeguards-to-prevent-misuse-and-will-fall-back-to-opus-4-8-for-high-risk-queries</link>
                                                                            <description>
                            <![CDATA[ The launch of Claude Fable 5 marks the first public release of a Mythos-class AI model ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">H6QQVm3bnJovuABuAajkNS</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/wjYii4ZLQmoYATphqrEovk-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 09 Jun 2026 17:04:20 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:48 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/wjYii4ZLQmoYATphqrEovk-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Anthropic Claude AI logo pictured on a smartphone screen, with AI tool branding blurred in background.]]></media:description>                                                            <media:text><![CDATA[Anthropic Claude AI logo pictured on a smartphone screen, with AI tool branding blurred in background.]]></media:text>
                                <media:title type="plain"><![CDATA[Anthropic Claude AI logo pictured on a smartphone screen, with AI tool branding blurred in background.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/wjYii4ZLQmoYATphqrEovk-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Anthropic has unveiled a new Mythos-class AI model designed specifically for public use. </p><p>The new model, dubbed Fable 5, boasts significant performance improvements on previous models, with the company noting that it “exceeds every model” previously made available to the public.</p><p>Notably, Fable 5 has been tweaked to include new safeguards that “block responses in specific high-risk areas” to prevent misuse – a key concern with the <a href="https://www.itpro.com/technology/artificial-intelligence/project-glasswing-anthropic-announces-big-tech-consortium-to-test-claude-mythos-ai-model-that-could-reshape-cybersecurity">launch of Claude Mythos</a>. </p><p>Claude Mythos was rolled-out as part of a gated release in April this year, with industry partners such as Microsoft, Amazon, and Cisco selected to test the model’s capabilities behind closed doors. </p><p>With the launch of Fable 5, Anthropic revealed that the model will “fall back” to Claude Opus 4.8 in the event that potential high-risk uses are detected. </p><p>Alongside the launch of Fable 5, Anthropic also revealed it will release Claude Mythos 5, again in a restricted capacity. </p><p>“Beginning today, all users who currently have access to Claude Mythos Preview (for example, our cybersecurity partners in <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-lets-glasswing-partners-publicly-share-mythos-flaws">Project Glasswing</a>) will be able to upgrade to Claude Mythos 5,” the company said in an announcement. </p><p>The company added that it plans to expand access in due course through a “more systematic trusted-access program”. </p><h2 id="performance-boosts-with-claude-fable-5">Performance boosts with Claude Fable 5</h2><p>According to Anthropic, Fable 5 shows “exceptional performance” in software engineering-related tasks, as well as knowledge work and scientific research. </p><p>Testing of the model by analytics company Hex revealed the model was the first to record a score of over 90% on its core analytics benchmark, which consists of running the model through “complex, long-running analytical tasks”. </p><p>Indeed, Hex noted that Fable recorded a 10-point improvement over Opus. </p><p>Similar testing by agentic workspace platform, Genspark, also showed significant performance boosts compared to other models, particularly in areas such as coding. </p><p>“Fable came out #1 on our evals, winning head-to-head against every model we tested,” the company said. “It was significantly stronger on the hardest tasks in the set — UI design and game coding."</p><h2 id="robust-guardrails-for-fable-5">Robust guardrails for Fable 5</h2><p>With Fable 5’s new safeguards, the model will be restricted from performing tasks in a range of areas, including cybersecurity and biology, and will revert back to Opus 4.8 to provide answers to queries on these topics. </p><p>“Releasing a model this capable comes with risks,” Anthropic said in a statement. “Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage.”</p><p>Anthropic said it “extensively red-teamed” classifiers to test the model’s defenses against potential jailbreaks as part of a bug bounty. </p><p>“Internally, we ran an external bug bounty that produced no universal jailbreaks in over 1,000 hours of testing,” the company said. “We then worked with external red-teaming orgs which also failed to find universal jailbreaks.”</p><h2 id="fable-5-pricing-and-access">Fable 5 pricing and access</h2><p>According to Anthropic, pricing for both models will stand at $10 per million input tokens, with $50 per million output tokens. </p><p>Token consumption costs have become a hot topic in the tech industry over the last two months with the rise of ‘tokenmaxxing’. </p><p>Uber, for example, revealed it <a href="https://www.itpro.com/technology/artificial-intelligence/ubers-eye-watering-ai-bill-shows-enterprises-are-still-measuring-ai-success-through-consumption-rather-than-outcomes-and-its-warping-our-perception-of-roi-and-productivity">blew through its entire annual AI budget in just four months</a> amidst skyrocketing use by staff. </p><p>Developers will be able to use Fable 5 through the Claude API while Mythos 5 will only be accessible to Project Glasswing participants. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ The AI rollback nobody wants to talk about... ]]></title>
                                                                                                <dc:content><![CDATA[ <p>From the outside, enterprise AI still looks like a rollout story in 2026: Copilots are being pushed across workforces, agents are moving from demos into production, and vendors are selling the idea that the next phase of AI adoption is mainly a question of scale.</p><p>Inside enterprises, the picture is a little messier. </p><p>Some projects are still moving ahead, but others are being trimmed, paused, or quietly dropped as CIOs take a harder look at cost, usage, risk, and measurable value when the dust and excitement settle. </p><p>To make sense of what is happening, <em>ITPro</em> spoke to Gartner analyst <a href="https://www.gartner.com/en/experts/anushree-verma"><u>Anushree Verma</u></a>, who observes that enterprise interest in copilots and agents is still accelerating, but that deployments without tangible value will now become much harder to sustain.</p><p>This situation leaves businesses with a more awkward question than whether they should adopt AI: They now have to work out which parts of their AI estate are worth keeping, which need more time, and which should be cut before they become another expensive layer. </p><h2 id="ai-pruning-not-abandonment">AI pruning, not abandonment</h2><p>The easy version of the story is that enterprises are rolling back from AI, but that is not quite right. In many cases, they are trying to do the opposite: move copilots and agents out of proof-of-concept (POC) work and into production.</p><p>The difference is that AI projects are being asked to survive normal enterprise scrutiny. </p><p>As Verma says, “I see a more accelerated interest in scaling copilots and agents into production in 2026. Most of the organizations have done at least one POC by the end of 2025, and they want to start getting value in their deployments. The deployments that do not show a tangible value will be difficult to sustain.”</p><p>Enterprises may still be expanding their AI use, but the bar is getting higher for tools that were initially approved on the promise of productivity gains, competitive pressure, employee excitement, or simple fear of being left behind.</p><h2 id="why-some-pilots-fail">Why (some) pilots fail</h2><p>Some AI projects never make it past the pilot stage because the technology is not the only thing being tested. On a basic level, the pilot also tests whether the business has a clear use case, clean enough data, proper controls, a willing user base, and a route from experimentation to measurable value.</p><p>Verma puts the problem down to “a combination of unclear business value, escalating cost, and inadequate risk controls.” Those three pressures tend to become more visible as projects move closer to production.</p><p>For example, a small team trial can absorb some rough edges, manual oversight, and unclear ownership, but a wider deployment needs deeper security rules, integration work, compliance checks, and so on. </p><p>A pilot is where the rollback becomes real: AI may look useful in a demo or limited team trial, but production is less forgiving. Once the full operating model comes into view, some deployments start to look less like transformation and more like expensive experiments.</p><h2 id="the-per-seat-copilot-problem">The per-seat copilot problem</h2><p>The clearest pressure point may be the broad, per-seat copilot. These types of AI tools are easy to understand and easy to roll out, but much harder to justify once the finance team starts asking what has changed.</p><p>A general-purpose AI copilot can help employees summarise meetings, draft documents, or search for information, but the problem is that these gains are often diffuse. In other words, the financial savings are hard to capture unless the business can turn it into lower costs, higher output, faster service, or better decisions.</p><p>When asked about this, Verma draws a sharp distinction: “[The] use cases with copilots are not leading to a tangible ROI. That kind of a use case gives a return on employees, while a workflow-specific AI tool gives an ROI.” </p><p>This phenomenon helps explain why narrower AI deployments may be easier to defend: A tool embedded in service management, compliance review, software testing, and so on has a clearer before-and-after test. </p><p>Simply put, it either cuts handling time, reduces manual work, improves accuracy, or clears a backlog – or it does not.</p><h2 id="agents-raise-the-stakes">Agents raise the stakes</h2><p>Agents sharpen the calculation because they promise more than assistance. A copilot can draft, summarise, or suggest, while an agent is supposed to complete tasks, trigger actions, and move work across systems with less human intervention.</p><p>The potential reward is larger, but so is the cost of poor design. </p><p>A narrowly scoped agent handling a repeatable internal workflow might be relatively easy to govern; a poorly defined agent with broad permissions, unclear ownership, or weak audit trails can create operational risk before it delivers much value.</p><p>As agentic AI moves out of experimentation, enterprises will need firmer boundaries around each deployment. CIOs, CISOs, and business owners need to know what the agent can do, which systems it can access, when a human needs to approve an action, and who is accountable if and when something breaks.</p><h2 id="governance-and-sprawl">Governance and sprawl</h2><p>Cost is only one reason for a quiet rollback. Governance may prove just as important, especially as enterprises move from a handful of AI experiments to a growing estate of copilots, agents, and workflow-specific tools.</p><p>Verma warns that businesses need “careful management of the agents deployed”, because “all the siloed, ad hoc task-specific agents are leading to an agent sprawl.”</p><p>The risk is that organizations end up with hundreds or thousands of small AI systems operating across departments, each with its own permissions, data access, owner, and failure modes.</p><p>At that point, an AI rollback starts to look less like caution and more like basic IT hygiene. </p><p>The scale of that challenge could grow quickly. Gartner has predicted that by 2028, “an average global Fortune 500 enterprise will have over 150,000 agents in use, up from less than 15 in 2025, generating significant agent sprawl, IT complexity and management challenges.”</p><h2 id="what-survives-post-correction">What survives post-correction?</h2><p>The deployments most likely to survive this correction will not necessarily be the most ambitious, but the ones attached to a clear workflow, a named business owner, a manageable risk profile, and a result the organization can actually measure.</p><p>“The use of agentic AI demands a careful assessment of use cases that can sustain value,” Verma says. “Each identified use case and business case must be evaluated on impact and return. They reflect different scales of investment, different risk tolerance, and operational complexity.”</p><p>Verma suggests a strategy that separates AI projects into three broad camps: defensive use cases that improve workforce productivity and operational efficiency, extension use cases that support growth or service differentiation, and more disruptive bets aimed at longer-term innovation.</p><p>Verma’s framework gives enterprises a more useful test than whether an AI project sounds impressive. Some deployments may deserve more time because they are improving how work gets done, while others may justify investment because they open up a new service, product, or operating model. </p><p>The weakest candidates are the ones that sit in the middle: costly to run, hard to govern, and too vague to show what they are changing.</p><p>The awkward but necessary work is knowing the difference. In 2026, the enterprises that look most serious about AI may not be the ones launching the most copilots or agents; they may be the ones willing to cut the deployments that no longer deserve a place in the stack.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/the-ai-rollback-nobody-wants-to-talk-about</link>
                                                                            <description>
                            <![CDATA[ When enterprises quietly scale back copilots and agents ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">Hyd3dE3EcY3tRyShsKPwYU</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/aKLAULJBFmuT2Aadzf9agi-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 05 Jun 2026 07:00:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:30:52 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Max Slater-Robins ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/aKLAULJBFmuT2Aadzf9agi-1920-80.jpg">
                                                            <media:credit><![CDATA[null]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Artificial Intelligence]]></media:description>                                                            <media:text><![CDATA[Artificial Intelligence]]></media:text>
                                <media:title type="plain"><![CDATA[Artificial Intelligence]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/aKLAULJBFmuT2Aadzf9agi-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>From the outside, enterprise AI still looks like a rollout story in 2026: Copilots are being pushed across workforces, agents are moving from demos into production, and vendors are selling the idea that the next phase of AI adoption is mainly a question of scale.</p><p>Inside enterprises, the picture is a little messier. </p><p>Some projects are still moving ahead, but others are being trimmed, paused, or quietly dropped as CIOs take a harder look at cost, usage, risk, and measurable value when the dust and excitement settle. </p><p>To make sense of what is happening, <em>ITPro</em> spoke to Gartner analyst <a href="https://www.gartner.com/en/experts/anushree-verma"><u>Anushree Verma</u></a>, who observes that enterprise interest in copilots and agents is still accelerating, but that deployments without tangible value will now become much harder to sustain.</p><p>This situation leaves businesses with a more awkward question than whether they should adopt AI: They now have to work out which parts of their AI estate are worth keeping, which need more time, and which should be cut before they become another expensive layer. </p><h2 id="ai-pruning-not-abandonment">AI pruning, not abandonment</h2><p>The easy version of the story is that enterprises are rolling back from AI, but that is not quite right. In many cases, they are trying to do the opposite: move copilots and agents out of proof-of-concept (POC) work and into production.</p><p>The difference is that AI projects are being asked to survive normal enterprise scrutiny. </p><p>As Verma says, “I see a more accelerated interest in scaling copilots and agents into production in 2026. Most of the organizations have done at least one POC by the end of 2025, and they want to start getting value in their deployments. The deployments that do not show a tangible value will be difficult to sustain.”</p><p>Enterprises may still be expanding their AI use, but the bar is getting higher for tools that were initially approved on the promise of productivity gains, competitive pressure, employee excitement, or simple fear of being left behind.</p><h2 id="why-some-pilots-fail">Why (some) pilots fail</h2><p>Some AI projects never make it past the pilot stage because the technology is not the only thing being tested. On a basic level, the pilot also tests whether the business has a clear use case, clean enough data, proper controls, a willing user base, and a route from experimentation to measurable value.</p><p>Verma puts the problem down to “a combination of unclear business value, escalating cost, and inadequate risk controls.” Those three pressures tend to become more visible as projects move closer to production.</p><p>For example, a small team trial can absorb some rough edges, manual oversight, and unclear ownership, but a wider deployment needs deeper security rules, integration work, compliance checks, and so on. </p><p>A pilot is where the rollback becomes real: AI may look useful in a demo or limited team trial, but production is less forgiving. Once the full operating model comes into view, some deployments start to look less like transformation and more like expensive experiments.</p><h2 id="the-per-seat-copilot-problem">The per-seat copilot problem</h2><p>The clearest pressure point may be the broad, per-seat copilot. These types of AI tools are easy to understand and easy to roll out, but much harder to justify once the finance team starts asking what has changed.</p><p>A general-purpose AI copilot can help employees summarise meetings, draft documents, or search for information, but the problem is that these gains are often diffuse. In other words, the financial savings are hard to capture unless the business can turn it into lower costs, higher output, faster service, or better decisions.</p><p>When asked about this, Verma draws a sharp distinction: “[The] use cases with copilots are not leading to a tangible ROI. That kind of a use case gives a return on employees, while a workflow-specific AI tool gives an ROI.” </p><p>This phenomenon helps explain why narrower AI deployments may be easier to defend: A tool embedded in service management, compliance review, software testing, and so on has a clearer before-and-after test. </p><p>Simply put, it either cuts handling time, reduces manual work, improves accuracy, or clears a backlog – or it does not.</p><h2 id="agents-raise-the-stakes">Agents raise the stakes</h2><p>Agents sharpen the calculation because they promise more than assistance. A copilot can draft, summarise, or suggest, while an agent is supposed to complete tasks, trigger actions, and move work across systems with less human intervention.</p><p>The potential reward is larger, but so is the cost of poor design. </p><p>A narrowly scoped agent handling a repeatable internal workflow might be relatively easy to govern; a poorly defined agent with broad permissions, unclear ownership, or weak audit trails can create operational risk before it delivers much value.</p><p>As agentic AI moves out of experimentation, enterprises will need firmer boundaries around each deployment. CIOs, CISOs, and business owners need to know what the agent can do, which systems it can access, when a human needs to approve an action, and who is accountable if and when something breaks.</p><h2 id="governance-and-sprawl">Governance and sprawl</h2><p>Cost is only one reason for a quiet rollback. Governance may prove just as important, especially as enterprises move from a handful of AI experiments to a growing estate of copilots, agents, and workflow-specific tools.</p><p>Verma warns that businesses need “careful management of the agents deployed”, because “all the siloed, ad hoc task-specific agents are leading to an agent sprawl.”</p><p>The risk is that organizations end up with hundreds or thousands of small AI systems operating across departments, each with its own permissions, data access, owner, and failure modes.</p><p>At that point, an AI rollback starts to look less like caution and more like basic IT hygiene. </p><p>The scale of that challenge could grow quickly. Gartner has predicted that by 2028, “an average global Fortune 500 enterprise will have over 150,000 agents in use, up from less than 15 in 2025, generating significant agent sprawl, IT complexity and management challenges.”</p><h2 id="what-survives-post-correction">What survives post-correction?</h2><p>The deployments most likely to survive this correction will not necessarily be the most ambitious, but the ones attached to a clear workflow, a named business owner, a manageable risk profile, and a result the organization can actually measure.</p><p>“The use of agentic AI demands a careful assessment of use cases that can sustain value,” Verma says. “Each identified use case and business case must be evaluated on impact and return. They reflect different scales of investment, different risk tolerance, and operational complexity.”</p><p>Verma suggests a strategy that separates AI projects into three broad camps: defensive use cases that improve workforce productivity and operational efficiency, extension use cases that support growth or service differentiation, and more disruptive bets aimed at longer-term innovation.</p><p>Verma’s framework gives enterprises a more useful test than whether an AI project sounds impressive. Some deployments may deserve more time because they are improving how work gets done, while others may justify investment because they open up a new service, product, or operating model. </p><p>The weakest candidates are the ones that sit in the middle: costly to run, hard to govern, and too vague to show what they are changing.</p><p>The awkward but necessary work is knowing the difference. In 2026, the enterprises that look most serious about AI may not be the ones launching the most copilots or agents; they may be the ones willing to cut the deployments that no longer deserve a place in the stack.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Snowflake and Anthropic are teaming up to push AI projects from pilot to production ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Snowflake and Anthropic have announced an expansion of their strategic partnership aimed at driving broader AI adoption for enterprises. </p><p>According to the duo, the move comes as both report increased demand from enterprises looking to move AI from pilot projects into full production. </p><p>The initial partnership, <a href="https://www.itpro.com/technology/artificial-intelligence/snowflake-inks-usd200m-deal-with-anthropic-to-drive-agentic-ai-in-the-enterprise">signed in December last year</a>, saw Anthropic’s Claude models integrated within Snowflake’s Cortex AI platform. </p><p>This allowed customers to use Claude with Snowflake data through the platform, helping to streamline access, support agent deployments, and provide users with a range of safety and governance controls. </p><p>So far, the partnership has proved beneficial for customers and highlights a growing demand for AI that “works directly on their governed data, not in isolated systems”, according to Christian Kleinerman, Snowflake’s EVP of product.</p><p>“We’re seeing strong demand across our AI products, with Snowflake Cortex Code becoming the fastest growing product in Snowflake’s history,” he said. </p><p>“Together with Anthropic, we’re helping organisations move from experimentation to product faster and laying the foundation for the agentic enterprises, where AI, data, and governance work together to drive real business outcomes.”</p><h2 id="snowflake-and-anthropic-target-deeper-integration">Snowflake and Anthropic target deeper integration</h2><p>Building on this, the duo has expanded access to Claude within the Cortex AI platform, now enabling enterprises to build AI agents and applications directly on in-house data. </p><p>The Cortex Agents framework, for example, allows customers to build custom agents spanning a range of use-cases, including customer support, data analysis, and core operations. </p><p>Elsewhere, Claude will also now power Snowflake Cortex Code and Snowflake Intelligence, the companies confirmed. </p><p>According to Snowflake, a host of customers, including Block, have recorded marked improvements using Claude across its core platforms. </p><p>Block engineering lead Arnaud Weber said the company has built an “AI-native operating layer” which connects data across different brand ecosystems, spanning CashApp, Afterpay, and Square. </p><p>Weber noted this has helped bolster security, data governance, and developer efficiency. </p><p>“By combining Anthropic Claude with Snowflake’s governed data platform, our teams can investigate compliance and security issues in real-time, trace controls and requirements, surface operational insights, and automate workflows grounded in trusted enterprise data,” Weber commented. </p><p>“Developers are also using Snowflake Cortex Code to build and operationalize these capabilities directly within Snowflake, creating a unified layer where AI can move seamlessly from analysis to action. </p><p>“This approach helps us reduce friction across investigations and decision-making, while maintaining the governance, performance, and scalability needed to apply AI responsibly across financial services and commerce.”</p><h2 id="moving-from-pilot-to-production">Moving from pilot to production</h2><p>The continued collaboration between Anthropic and Snowflake targets a recurring pain point for enterprises - mainly the inability to move projects out of the pilot stage and into full production. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence/uk-firms-are-stuck-in-an-ai-pilot-purgatory-heres-how-they-can-drive-success"><u>Research from Google</u></a> as far back as 2024 identified this as a key challenge for enterprises dabbling in AI adoption, describing the trend as “pilot purgatory”.</p><p>Crucially, the situation has remained much the same over the last two years. <a href="https://www.itpro.com/technology/artificial-intelligence/only-a-handful-of-generative-ai-projects-make-it-into-production-heres-why"><u>Analysis from Informatica </u></a>in January 2025 found just over one-third (38%) of companies had successfully transitioned AI projects into full production, for example. </p><p>This problem has been further exacerbated with the advent of agentic AI, with integration of agents proving far more complicated. Research from Dynatrace in January this year, for example, found around <a href="https://www.itpro.com/technology/artificial-intelligence/half-of-agentic-ai-projects-are-still-stuck-at-the-pilot-stage-but-thats-not-stopping-enterprises-from-ramping-up-investment">half of AI agent projects are still stuck in proof of concept (PoC) stages</a>. </p><p>A key factor here, the study noted, lies in the fact enterprises fail to efficiently govern agents, which prevents them from scaling projects safely. More than half (52%) of IT leaders highlighted concerns around security, privacy, and compliance as key hurdles. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/business/data-and-insights/snowflake-and-anthropic-are-teaming-up-to-push-ai-projects-from-pilot-to-production</link>
                                                                            <description>
                            <![CDATA[ The expanded collaboration looks to solve a recurring pain point for enterprises ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">Z89o6Qde9krmZak3iJ8xMK</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/4Vpt5mudMjnUa9KHDfEJyX-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 03 Jun 2026 10:43:28 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:19 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Data and Insights]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                    <category><![CDATA[Business]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/4Vpt5mudMjnUa9KHDfEJyX-1920-80.jpg">
                                                            <media:credit><![CDATA[Snowflake/Anthropic]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Snowflake and Anthropic logos pictured side by side with aurora borealis lights in corners of image.]]></media:description>                                                            <media:text><![CDATA[Snowflake and Anthropic logos pictured side by side with aurora borealis lights in corners of image.]]></media:text>
                                <media:title type="plain"><![CDATA[Snowflake and Anthropic logos pictured side by side with aurora borealis lights in corners of image.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/4Vpt5mudMjnUa9KHDfEJyX-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Snowflake and Anthropic have announced an expansion of their strategic partnership aimed at driving broader AI adoption for enterprises. </p><p>According to the duo, the move comes as both report increased demand from enterprises looking to move AI from pilot projects into full production. </p><p>The initial partnership, <a href="https://www.itpro.com/technology/artificial-intelligence/snowflake-inks-usd200m-deal-with-anthropic-to-drive-agentic-ai-in-the-enterprise">signed in December last year</a>, saw Anthropic’s Claude models integrated within Snowflake’s Cortex AI platform. </p><p>This allowed customers to use Claude with Snowflake data through the platform, helping to streamline access, support agent deployments, and provide users with a range of safety and governance controls. </p><p>So far, the partnership has proved beneficial for customers and highlights a growing demand for AI that “works directly on their governed data, not in isolated systems”, according to Christian Kleinerman, Snowflake’s EVP of product.</p><p>“We’re seeing strong demand across our AI products, with Snowflake Cortex Code becoming the fastest growing product in Snowflake’s history,” he said. </p><p>“Together with Anthropic, we’re helping organisations move from experimentation to product faster and laying the foundation for the agentic enterprises, where AI, data, and governance work together to drive real business outcomes.”</p><h2 id="snowflake-and-anthropic-target-deeper-integration">Snowflake and Anthropic target deeper integration</h2><p>Building on this, the duo has expanded access to Claude within the Cortex AI platform, now enabling enterprises to build AI agents and applications directly on in-house data. </p><p>The Cortex Agents framework, for example, allows customers to build custom agents spanning a range of use-cases, including customer support, data analysis, and core operations. </p><p>Elsewhere, Claude will also now power Snowflake Cortex Code and Snowflake Intelligence, the companies confirmed. </p><p>According to Snowflake, a host of customers, including Block, have recorded marked improvements using Claude across its core platforms. </p><p>Block engineering lead Arnaud Weber said the company has built an “AI-native operating layer” which connects data across different brand ecosystems, spanning CashApp, Afterpay, and Square. </p><p>Weber noted this has helped bolster security, data governance, and developer efficiency. </p><p>“By combining Anthropic Claude with Snowflake’s governed data platform, our teams can investigate compliance and security issues in real-time, trace controls and requirements, surface operational insights, and automate workflows grounded in trusted enterprise data,” Weber commented. </p><p>“Developers are also using Snowflake Cortex Code to build and operationalize these capabilities directly within Snowflake, creating a unified layer where AI can move seamlessly from analysis to action. </p><p>“This approach helps us reduce friction across investigations and decision-making, while maintaining the governance, performance, and scalability needed to apply AI responsibly across financial services and commerce.”</p><h2 id="moving-from-pilot-to-production">Moving from pilot to production</h2><p>The continued collaboration between Anthropic and Snowflake targets a recurring pain point for enterprises - mainly the inability to move projects out of the pilot stage and into full production. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence/uk-firms-are-stuck-in-an-ai-pilot-purgatory-heres-how-they-can-drive-success"><u>Research from Google</u></a> as far back as 2024 identified this as a key challenge for enterprises dabbling in AI adoption, describing the trend as “pilot purgatory”.</p><p>Crucially, the situation has remained much the same over the last two years. <a href="https://www.itpro.com/technology/artificial-intelligence/only-a-handful-of-generative-ai-projects-make-it-into-production-heres-why"><u>Analysis from Informatica </u></a>in January 2025 found just over one-third (38%) of companies had successfully transitioned AI projects into full production, for example. </p><p>This problem has been further exacerbated with the advent of agentic AI, with integration of agents proving far more complicated. Research from Dynatrace in January this year, for example, found around <a href="https://www.itpro.com/technology/artificial-intelligence/half-of-agentic-ai-projects-are-still-stuck-at-the-pilot-stage-but-thats-not-stopping-enterprises-from-ramping-up-investment">half of AI agent projects are still stuck in proof of concept (PoC) stages</a>. </p><p>A key factor here, the study noted, lies in the fact enterprises fail to efficiently govern agents, which prevents them from scaling projects safely. More than half (52%) of IT leaders highlighted concerns around security, privacy, and compliance as key hurdles. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Fujitsu partners with Anthropic to drive AI transformation in Japan ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Fujitsu has announced a strategic partnership with Anthropic designed to accelerate <a href="https://www.itpro.com/technology/artificial-intelligence/ai-adoption-is-finally-driving-roi-for-b2b-teams-in-the-uk-and-eu">AI adoption</a> across Japanese enterprises and strengthen cybersecurity for critical infrastructure.</p><p>The agreement will see Anthropic's AI technologies – including <a href="https://www.itpro.com/software/development/anthropic-claude-code-usage-limits-increase-spacex-compute-deal">Claude</a> – paired with Fujitsu's industry expertise in delivering mission-critical systems to improve AI transformation capabilities across sectors such as government, finance, healthcare, defense, and critical infrastructure.</p><p>As part of the initiative, approximately 100,000 Fujitsu Group employees will utilize Claude to enhance operations, accelerate development processes, and help establish frameworks for the safe and reliable use of AI.</p><p>The company said the knowledge gained from this internal deployment will then be used to help customers accelerate their own AI adoption efforts, while ensuring compliance, security, and operational reliability.</p><p>The partnership will also see Fujitsu gain early access to <a href="https://www.itpro.com/software/development/anthropic-claude-opus-4-software-development">Anthropic's latest AI models</a>, with a view to incorporating the technology into future customer solutions and services.</p><p>In an announcement, Takahito Tokita, representative director and CEO at Fujitsu, said the company sees the evolution of AI as something that must be swiftly implemented across society and translated into real business value.</p><p>"Through this collaboration, we will combine Fujitsu's deep expertise across industries and business functions – particularly its extensive know-how in mission-critical domains – with Anthropic's advanced AI models," Tokita said. "In doing so, we aim to support the creation of new value across industries and realize a trustworthy, AI-driven society."</p><h2 id="expansion-of-the-fujitsu-fde-model">Expansion of the Fujitsu FDE model</h2><p>At the heart of the partnership is an expansion of Fujitsu's Forward Deployed Engineer (FDE) model, which focuses on working closely with customers to translate AI capabilities into business outcomes.</p><p>The firm said it will now combine Claude with its own AI technologies implemented as part of the initiative, which include the Fujitsu Kozuchi AI platform as well as its Takane LLM.</p><p>According to Fujitsu, this approach will allow it to "go beyond simple AI deployment" and deliver AI offerings tailored to customers' specific data sovereignty, regulatory compliance, security, and performance needs.</p><p>"Through this partnership, we will further strengthen and accelerate our FDE model, ensuring that AI is continuously translated into real value through deep engagement with customer operations," explained Yoshinami Takahashi, Fujitsu's corporate executive officer, corporate vice president, and COO in charge of solution services.</p><p>"This will accelerate structural transformation of business and enable a shift toward high-value-added business models."</p><h2 id="strengthening-cyber-defense-capabilities">Strengthening cyber defense capabilities</h2><p>Alongside AI adoption, the companies said the partnership will also place a focus on bolstering cyber defense capabilities as organizations increasingly look to secure their critical systems in the era of AI.</p><p>Fujitsu said it intends to move beyond traditional expert-led cyber security models by combining human expertise with AI-assisted operations to improve response times and strengthen protection across enterprises, essential services, and critical infrastructure.</p><p>The company added that it will work collaboratively with the Japanese government to apply the lessons learned from the initiative to improve security across wider society.</p><p>Commenting on the partnership, Anthropic chief commercial officer Paul Smith described the agreement as one of the most significant AI commitments in the Japanese market to date.</p><p>"The institutions that anchor Japanese society – its banks, its hospitals, its government, its critical infrastructure – hold AI to the highest standard," he said. "Fujitsu has been the technology partner to those institutions for decades, and they are now deploying Claude to 100,000 of their own employees and building a 1,000-person engineering team to bring it to their customers.</p><p>"This is one of the most consequential commitments to frontier AI in the Japanese market, and we're proud for Anthropic to be the partner Fujitsu trusts to deliver on that commitment." </p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/fujitsu-partners-with-anthropic-to-drive-ai-transformation-in-japan</link>
                                                                            <description>
                            <![CDATA[ The agreement will see the technology giant deploy Claude across its workforce while strengthening AI services and security capabilities for customers ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">7Mc76rUTKzKc8A2mfYddR3</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/i2oAS8c6EvmAkEs3hUnhPa-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 28 May 2026 08:13:18 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:30 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ itpro@futurenet.com (Daniel Todd) ]]></author>                    <dc:creator><![CDATA[ Daniel Todd ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/SRyC34qeLpNDj3dJtsVDhT-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/i2oAS8c6EvmAkEs3hUnhPa-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[ general view of sign outside a Fujitsu office on January 15, 2024 in Warrington, United Kingdom]]></media:description>                                                            <media:text><![CDATA[ general view of sign outside a Fujitsu office on January 15, 2024 in Warrington, United Kingdom]]></media:text>
                                <media:title type="plain"><![CDATA[ general view of sign outside a Fujitsu office on January 15, 2024 in Warrington, United Kingdom]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/i2oAS8c6EvmAkEs3hUnhPa-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Fujitsu has announced a strategic partnership with Anthropic designed to accelerate <a href="https://www.itpro.com/technology/artificial-intelligence/ai-adoption-is-finally-driving-roi-for-b2b-teams-in-the-uk-and-eu">AI adoption</a> across Japanese enterprises and strengthen cybersecurity for critical infrastructure.</p><p>The agreement will see Anthropic's AI technologies – including <a href="https://www.itpro.com/software/development/anthropic-claude-code-usage-limits-increase-spacex-compute-deal">Claude</a> – paired with Fujitsu's industry expertise in delivering mission-critical systems to improve AI transformation capabilities across sectors such as government, finance, healthcare, defense, and critical infrastructure.</p><p>As part of the initiative, approximately 100,000 Fujitsu Group employees will utilize Claude to enhance operations, accelerate development processes, and help establish frameworks for the safe and reliable use of AI.</p><p>The company said the knowledge gained from this internal deployment will then be used to help customers accelerate their own AI adoption efforts, while ensuring compliance, security, and operational reliability.</p><p>The partnership will also see Fujitsu gain early access to <a href="https://www.itpro.com/software/development/anthropic-claude-opus-4-software-development">Anthropic's latest AI models</a>, with a view to incorporating the technology into future customer solutions and services.</p><p>In an announcement, Takahito Tokita, representative director and CEO at Fujitsu, said the company sees the evolution of AI as something that must be swiftly implemented across society and translated into real business value.</p><p>"Through this collaboration, we will combine Fujitsu's deep expertise across industries and business functions – particularly its extensive know-how in mission-critical domains – with Anthropic's advanced AI models," Tokita said. "In doing so, we aim to support the creation of new value across industries and realize a trustworthy, AI-driven society."</p><h2 id="expansion-of-the-fujitsu-fde-model">Expansion of the Fujitsu FDE model</h2><p>At the heart of the partnership is an expansion of Fujitsu's Forward Deployed Engineer (FDE) model, which focuses on working closely with customers to translate AI capabilities into business outcomes.</p><p>The firm said it will now combine Claude with its own AI technologies implemented as part of the initiative, which include the Fujitsu Kozuchi AI platform as well as its Takane LLM.</p><p>According to Fujitsu, this approach will allow it to "go beyond simple AI deployment" and deliver AI offerings tailored to customers' specific data sovereignty, regulatory compliance, security, and performance needs.</p><p>"Through this partnership, we will further strengthen and accelerate our FDE model, ensuring that AI is continuously translated into real value through deep engagement with customer operations," explained Yoshinami Takahashi, Fujitsu's corporate executive officer, corporate vice president, and COO in charge of solution services.</p><p>"This will accelerate structural transformation of business and enable a shift toward high-value-added business models."</p><h2 id="strengthening-cyber-defense-capabilities">Strengthening cyber defense capabilities</h2><p>Alongside AI adoption, the companies said the partnership will also place a focus on bolstering cyber defense capabilities as organizations increasingly look to secure their critical systems in the era of AI.</p><p>Fujitsu said it intends to move beyond traditional expert-led cyber security models by combining human expertise with AI-assisted operations to improve response times and strengthen protection across enterprises, essential services, and critical infrastructure.</p><p>The company added that it will work collaboratively with the Japanese government to apply the lessons learned from the initiative to improve security across wider society.</p><p>Commenting on the partnership, Anthropic chief commercial officer Paul Smith described the agreement as one of the most significant AI commitments in the Japanese market to date.</p><p>"The institutions that anchor Japanese society – its banks, its hospitals, its government, its critical infrastructure – hold AI to the highest standard," he said. "Fujitsu has been the technology partner to those institutions for decades, and they are now deploying Claude to 100,000 of their own employees and building a 1,000-person engineering team to bring it to their customers.</p><p>"This is one of the most consequential commitments to frontier AI in the Japanese market, and we're proud for Anthropic to be the partner Fujitsu trusts to deliver on that commitment." </p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Are AI tools making us less intelligent?  ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Are we diluting our cognitive abilities by looking for answers through algorithms and machines?  </p><p>It's a question that is gaining prominence in our rapidly advancing digital world. <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-get-used-to-working-with-ai-powered-digital-colleagues">AI-powered technologies</a> provide instant answers or solutions, undoubtedly making us more efficient and helping businesses do more than ever before. But there are growing concerns that this is at the expense of our ability to think for ourselves. </p><p>Experts and workers are now asking, "Will AI make us less intelligent?"</p><p>The Royal Observatory Greenwich is concerned it will. Indeed, it has warned about the need to avoid "complete dependence" on AI. Paddy Rodgers, the director of the Royal Museums Greenwich group, told the <a href="https://www.bbc.co.uk/news/articles/c2023l60370o"><em>BBC</em></a> that "a reliance solely on <a href="https://www.itpro.com/networking/27171/what-is-a-chatbot">instant answers</a> risks losing the habits of questioning and evaluation that underpin knowledge, expertise, and innovation". </p><p>Workers are also worried, according to a study by SaaS business GoTo (formerly LogMeIn). Its report, <a href="https://www.goto.com/resources/pulse-of-work-2026"><em>The Pulse of Work in 2026: Opportunity, Risk, and Responsibility in an AI-Driven Workplace</em></a>, takes into account the opinion of 2,500 employees and IT leaders from around the world. </p><p>The report states that 50% of employees say they <a href="https://www.itpro.com/technology/artificial-intelligence/ai-tools-critical-thinking-reliance">rely too much on AI</a>, while 30% say they feel they can't function without it. Crucially, more than a third  (39%) feel their reliance on AI tools is eroding their skills and making them less intelligent. </p><p>The issue is more acute among younger workers, with nearly half of Gen Z (under 30) reporting similar concerns that AI dependency is making them less intelligent. </p><p>One of the reasons for the overreliance on AI tools, according to GoTo's report, is the pressure to be more efficient, with 60% of employees reporting that they felt the need to <a href="https://www.itpro.com/software/development/ai-coding-tools-arent-the-solution-to-the-unfolding-developer-crisis-teams-think-they-can-boost-productivity-and-delivery-times-but-end-up-bogged-down-by-manual-remediation-and-unsafe-code">boost their productivity</a>. However, the report doesn't clarify exactly where that pressure comes from.  </p><p>"<a href="https://www.itpro.com/business/leadership/why-responsible-ai-development-relies-on-female-leadership-input">Responsible AI</a> use is about having the right tools and supporting the people who use them," said Dan Schawbel, managing partner at Workplace Intelligence, which helped conduct the research the report is based on. </p><p>"Our research highlights the importance of equipping employees with the skills, policies, and guidance they need to work alongside AI effectively. Companies that get this right will see productivity gains while ensuring their workforce remains confident, capable, and engaged."</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/are-ai-tools-making-us-less-intelligent</link>
                                                                            <description>
                            <![CDATA[ Growing reliance on AI tools that provide instant answers could be eroding skills, GoTo report warns ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">eSmFhi65giZvvTfxQMEtXS</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/XqgDmsuu64a3kStF7Pkbq7-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 19 May 2026 13:00:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:32:04 +0000</updated>
                                                                                                                                            <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ itpro@futurenet.com (Bobby Hellard) ]]></author>                    <dc:creator><![CDATA[ Bobby Hellard ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/bsR2tHSyVKUoyXZF5pNsDA-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Bobby Hellard&amp;nbsp;is&amp;nbsp;ITPro&#039;s Reviews Editor and has worked on&amp;nbsp;CloudPro and ChannelPro since 2018. In his time at ITPro, Bobby has covered stories for all the major technology companies, such as Apple, Microsoft, Amazon and Facebook, and regularly attends industry-leading events such as AWS Re:Invent and Google Cloud Next.&lt;/p&gt;
&lt;p&gt;Bobby mainly covers hardware reviews, but you will also recognize him as the face of many of our video reviews of laptops and smartphones.&lt;/p&gt;
&lt;p&gt;He has been a journalist for ten years, originally covering sports, before moving into business technology with ITPro. He has bylines in The Independent, Vice and The Business Briefing. Contact him at &lt;a href=&quot;mailto:bobby.hellard@futurenet.com&quot;&gt;bobby.hellard@futurenet.com&lt;/a&gt; or find him on Twitter: &lt;a href=&quot;https://twitter.com/bobbyhellard&quot;&gt;@bobbyhellard&lt;/a&gt;&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/XqgDmsuu64a3kStF7Pkbq7-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A collection of AI apps on a smartphone]]></media:description>                                                            <media:text><![CDATA[A collection of AI apps on a smartphone]]></media:text>
                                <media:title type="plain"><![CDATA[A collection of AI apps on a smartphone]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/XqgDmsuu64a3kStF7Pkbq7-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Are we diluting our cognitive abilities by looking for answers through algorithms and machines?  </p><p>It's a question that is gaining prominence in our rapidly advancing digital world. <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-get-used-to-working-with-ai-powered-digital-colleagues">AI-powered technologies</a> provide instant answers or solutions, undoubtedly making us more efficient and helping businesses do more than ever before. But there are growing concerns that this is at the expense of our ability to think for ourselves. </p><p>Experts and workers are now asking, "Will AI make us less intelligent?"</p><p>The Royal Observatory Greenwich is concerned it will. Indeed, it has warned about the need to avoid "complete dependence" on AI. Paddy Rodgers, the director of the Royal Museums Greenwich group, told the <a href="https://www.bbc.co.uk/news/articles/c2023l60370o"><em>BBC</em></a> that "a reliance solely on <a href="https://www.itpro.com/networking/27171/what-is-a-chatbot">instant answers</a> risks losing the habits of questioning and evaluation that underpin knowledge, expertise, and innovation". </p><p>Workers are also worried, according to a study by SaaS business GoTo (formerly LogMeIn). Its report, <a href="https://www.goto.com/resources/pulse-of-work-2026"><em>The Pulse of Work in 2026: Opportunity, Risk, and Responsibility in an AI-Driven Workplace</em></a>, takes into account the opinion of 2,500 employees and IT leaders from around the world. </p><p>The report states that 50% of employees say they <a href="https://www.itpro.com/technology/artificial-intelligence/ai-tools-critical-thinking-reliance">rely too much on AI</a>, while 30% say they feel they can't function without it. Crucially, more than a third  (39%) feel their reliance on AI tools is eroding their skills and making them less intelligent. </p><p>The issue is more acute among younger workers, with nearly half of Gen Z (under 30) reporting similar concerns that AI dependency is making them less intelligent. </p><p>One of the reasons for the overreliance on AI tools, according to GoTo's report, is the pressure to be more efficient, with 60% of employees reporting that they felt the need to <a href="https://www.itpro.com/software/development/ai-coding-tools-arent-the-solution-to-the-unfolding-developer-crisis-teams-think-they-can-boost-productivity-and-delivery-times-but-end-up-bogged-down-by-manual-remediation-and-unsafe-code">boost their productivity</a>. However, the report doesn't clarify exactly where that pressure comes from.  </p><p>"<a href="https://www.itpro.com/business/leadership/why-responsible-ai-development-relies-on-female-leadership-input">Responsible AI</a> use is about having the right tools and supporting the people who use them," said Dan Schawbel, managing partner at Workplace Intelligence, which helped conduct the research the report is based on. </p><p>"Our research highlights the importance of equipping employees with the skills, policies, and guidance they need to work alongside AI effectively. Companies that get this right will see productivity gains while ensuring their workforce remains confident, capable, and engaged."</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Workers are wasting a full work day each week switching between AI tools ]]></title>
                                                                                                <dc:content><![CDATA[ <p>British workers are wasting nearly a full work day each week switching between disconnected<a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today"> AI tools</a> and systems. </p><p>While more than half of employees say <a href="https://www.itpro.com/business/business-strategy/ai-is-boosting-personal-productivity-but-slowing-down-teams-heres-why">AI helps them complete individual tasks more efficiently</a>, any time saved is often wiped out by the time spent switching between systems, checking outputs, and manually moving information across tools.</p><p>In fact, <a href="https://forms.workday.com/en-gb/reports/copy-paste-economy-ai-research/form.html?locale=engb&amp;aud=br&amp;stage=at&amp;pblr=wdcmk&amp;camp=701VT00000pAGm5YAG&amp;assettype=rep&amp;productfocus=wds&amp;utm_source=pubre&amp;step=step1_default" target="_blank"><u>research</u></a> from enterprise AI platform Workday shows one-in-four UK workers spends seven or more hours a week copying information between applications, reconciling conflicting data, and manually feeding context into AI tools.</p><p>According to Workday, a key issue here is that businesses are prioritizing new features without ensuring that the technology can work together. While an isolated application might speed up one specific task, managing the disconnect slows down the rest of the day for workers.</p><p>"Too many employees are serving as the human middleware between disconnected AI systems," said Daniel Pell, vice president and country manager, UKI at Workday. </p><p>"The companies seeing the most value from AI are building it directly into the systems where their people, data and work come together."</p><p>Nearly eight-in-ten UK workers are held up by annoying administrative tasks such as chasing down data just to feed it into an <a href="https://www.itpro.com/technology/artificial-intelligence/google-boasts-that-a-single-gemini-prompt-uses-roughly-the-same-energy-as-a-basic-search-but-thats-not-painting-the-full-picture">AI prompt</a>. And while they're working harder to manage new tools, their output doesn't reflect this. </p><p>Three-quarters of workers said decisions are delayed when information is missing or unclear, while a similar number of teams often disagree over whose numbers are right.</p><p>"Dealing with system glitches, chasing approval and constantly fixing or redoing work because of inconsistent data – it keeps me busy, but doesn’t feel like real progress," one IT director told the researchers.</p><p>As a result, more than 60% of UK workers find themselves stuck in "busy but unproductive" days often or very often - many more than the global figure of 43%. </p><h2 id="confusion-fuels-burnout">Confusion fuels burnout</h2><p>According to Workday, this is causing serious issues with burnout. More than three-quarters (77%) of respondents reported higher levels of stress in their role due to navigating disparate AI tools and systems.</p><p>"My day often feels busy but not genuinely productive when I’m pulled into constant coordination tasks and system-related issues that interrupt focused, high-value work," said one respondent.</p><p>On a more positive note, 81% of UK employees said that AI is improving how their day-to-day work feels, and 45% said it has accelerated their work in a productive way. </p><p>However, only 23% of UK organizations have AI deeply embedded in their core systems, leaving it on the periphery. Among those organizations that do, 57% of employees report that AI has reduced task time by 25% or more. </p><p>Where AI isn't used in core systems at all, that figure is 39%, Workday found.</p><p>"Employees are already showing organizations how they want AI to work: integrated directly into workflows, proactively surfacing insights and handling coordination in the background," the researchers said.</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/too-many-employees-are-serving-as-the-human-middleware-workers-are-wasting-a-full-work-day-each-week-switching-between-disparate-ai-tools-and-internal-systems</link>
                                                                            <description>
                            <![CDATA[ Transferring data from one AI tool to another is costing more time than the tools actually save ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">rr2psx2usn5GvWS7by4zRM</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/qFLihzPEoexhv3ez2cxbsL-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 15 May 2026 09:35:55 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:31:14 +0000</updated>
                                                                                                                                            <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Emma Woollacott ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/aWfskavxoVSMDy6cDWtYmJ-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/qFLihzPEoexhv3ez2cxbsL-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Visibly stressed male office worker looking at desktop computer screen with clenched fists placed under chin, while female colleague works on laptop in background.]]></media:description>                                                            <media:text><![CDATA[Visibly stressed male office worker looking at desktop computer screen with clenched fists placed under chin, while female colleague works on laptop in background.]]></media:text>
                                <media:title type="plain"><![CDATA[Visibly stressed male office worker looking at desktop computer screen with clenched fists placed under chin, while female colleague works on laptop in background.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/qFLihzPEoexhv3ez2cxbsL-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>British workers are wasting nearly a full work day each week switching between disconnected<a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today"> AI tools</a> and systems. </p><p>While more than half of employees say <a href="https://www.itpro.com/business/business-strategy/ai-is-boosting-personal-productivity-but-slowing-down-teams-heres-why">AI helps them complete individual tasks more efficiently</a>, any time saved is often wiped out by the time spent switching between systems, checking outputs, and manually moving information across tools.</p><p>In fact, <a href="https://forms.workday.com/en-gb/reports/copy-paste-economy-ai-research/form.html?locale=engb&amp;aud=br&amp;stage=at&amp;pblr=wdcmk&amp;camp=701VT00000pAGm5YAG&amp;assettype=rep&amp;productfocus=wds&amp;utm_source=pubre&amp;step=step1_default" target="_blank"><u>research</u></a> from enterprise AI platform Workday shows one-in-four UK workers spends seven or more hours a week copying information between applications, reconciling conflicting data, and manually feeding context into AI tools.</p><p>According to Workday, a key issue here is that businesses are prioritizing new features without ensuring that the technology can work together. While an isolated application might speed up one specific task, managing the disconnect slows down the rest of the day for workers.</p><p>"Too many employees are serving as the human middleware between disconnected AI systems," said Daniel Pell, vice president and country manager, UKI at Workday. </p><p>"The companies seeing the most value from AI are building it directly into the systems where their people, data and work come together."</p><p>Nearly eight-in-ten UK workers are held up by annoying administrative tasks such as chasing down data just to feed it into an <a href="https://www.itpro.com/technology/artificial-intelligence/google-boasts-that-a-single-gemini-prompt-uses-roughly-the-same-energy-as-a-basic-search-but-thats-not-painting-the-full-picture">AI prompt</a>. And while they're working harder to manage new tools, their output doesn't reflect this. </p><p>Three-quarters of workers said decisions are delayed when information is missing or unclear, while a similar number of teams often disagree over whose numbers are right.</p><p>"Dealing with system glitches, chasing approval and constantly fixing or redoing work because of inconsistent data – it keeps me busy, but doesn’t feel like real progress," one IT director told the researchers.</p><p>As a result, more than 60% of UK workers find themselves stuck in "busy but unproductive" days often or very often - many more than the global figure of 43%. </p><h2 id="confusion-fuels-burnout">Confusion fuels burnout</h2><p>According to Workday, this is causing serious issues with burnout. More than three-quarters (77%) of respondents reported higher levels of stress in their role due to navigating disparate AI tools and systems.</p><p>"My day often feels busy but not genuinely productive when I’m pulled into constant coordination tasks and system-related issues that interrupt focused, high-value work," said one respondent.</p><p>On a more positive note, 81% of UK employees said that AI is improving how their day-to-day work feels, and 45% said it has accelerated their work in a productive way. </p><p>However, only 23% of UK organizations have AI deeply embedded in their core systems, leaving it on the periphery. Among those organizations that do, 57% of employees report that AI has reduced task time by 25% or more. </p><p>Where AI isn't used in core systems at all, that figure is 39%, Workday found.</p><p>"Employees are already showing organizations how they want AI to work: integrated directly into workflows, proactively surfacing insights and handling coordination in the background," the researchers said.</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ How Welsh councils are improving services with Microsoft Copilot ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Microsoft 365 Copilot is becoming widespread, as the go-to <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> service at firms already invested in the M365 productivity suite. The AI tool is the first port of call many organizations will have with generative AI and in the right hands, can drive meaningful increases in productivity.</p><p>Some standout examples of benefits can be found in the public sector, where pioneering councils are already using Copilot to the benefit of their staff.</p><p><em>ITPro</em> spoke to representatives to three such councils in Wales, who have led the charge on AI adoption in their respective roles: Mari Ropstad, head of service at Rhondda Cynon Taf County Borough Council; Ness Young, director of Corporate Service, City &amp; County of Swansea; and Georgia Sweet, digital transformation project officer, Carmarthenshire County Council.</p><h2 id="concrete-benefits">Concrete benefits</h2><p>Microsoft figures show that across the councils are already saving time and improving processes using Copilot.</p><p>Swansea County Council estimates that it saved 5,400 staff hours in just four weeks, as a direct effect of AI usage.</p><p>In Rhondda Cynon Taf County Borough Council, leaders have noted marked improvements in core tasks such as assessments, which Copilot has allowed to be created almost four times faster. After adopting the tool, 75% of staff say they feel more positive about their role, with 86% also reporting the quality of their work has improved.</p><p>At Carmarthenshire County Council, Sweet explains her team has developed a Copilot agent that responds to customer enquiries with FAQs drawn from the website. The deployment has helped the council to meet its ten-day target for complaints.</p><p>In addition to improving the customer-facing experience for Carmarthenshire County Council, Copilot has also brought benefits for workers inside the organization.</p><p>“We've seen a lot of feedback from <a href="https://www.itpro.com/business/careers-and-training/how-can-leaders-support-neurodiversity-in-tech"><u>neurodiverse</u></a> colleagues, particularly those with ADHD and dyslexia, where they're finding that Copilot is transforming the way that they work,” Sweet says.</p><p>“So they're finding it easy to process information, it's really helping with their written communication and the tools like rewrite or summarizing large documents, that's really helping them and improving their experience at work.”</p><p>Sweet adds that Copilot can be used without technical experience, with options to take it further depending on the user’s skill level.</p><h2 id="teething-difficulties">Teething difficulties</h2><p>Despite Copilot’s relative ease of access, users do need a basic understanding of Microsoft and AI systems to make the most of the tool – and for some staff, this was an initial stumbling block. </p><p>Young says that adoption of Copilot has revealed that some staff are less digitally literate than expected, necessitating additional time for training.</p><p>“We've got about 3,500 people using Copilot in Swansea, and some of the feedback that we've had from those providing training and engagement is that digital literacy can be an obstacle to initial use,” Young tells <em>ITPro</em>.</p><p>“Not necessarily just in a technical sense, but also in terms of confidence. And so we have invested heavily, we've got a support partner working with us to train people and we also have an engagement officer who provides one to one support particularly for those who have full licenses.”</p><p>Ropstad’s experience has been similar, with her team having found that users need “concrete use cases” for AI tools beyond basic training, including suggested prompts for generating forms or summarizing notes.</p><p>“So we use a lot of our own practitioners to teach other practitioners,” Ropstad explains. “And for us, it's always about trying to bring people into this middle band. You know, we are a council we can't afford for quality to be haphazard, it has to be a standard of quality. And so in order to get everybody to a middle band of Copilot use, we have to be able to bring everybody up to that level.”</p><p>All three councils benefited from first providing Copilot to a small group of ‘hero’ users, those most enthusiastic about AI and with the greatest technical experience, who helped to identify the most promising avenues for the technology and down the line to share their expertise with colleagues.</p><p>At Swansea Council, Young says 100 licenses were initially assigned, with 75 given to workers on pilot projects identified through discussions with Microsoft and other Welsh councils. The remaining 25 were assigned to the council’s senior leadership team, to make it a “leadership opportunity and really drive it from the top of the organization”, Young adds.</p><p>Carmarthenshire County Council started with a similarly small group, who later shared their learnings with the wider organization. Internally, employees are now sharing prompts via Teams, as well as ideas for AI use.</p><p>“I think people are so keen to share their learnings and if they found an amazing prompt, or if they've got a new process they're using Copilot for, and I think that's quite rare with new technology, to have that kind of buzz and excitement,” Sweet says.</p><p>“It’s as though the product sort of sells itself,” Young adds.</p><p>At Rhondda Cynon Taf County Borough Council, around 350 staff are now using Microsoft 365 Copilot. Initial concerns at the council centered around the ethical concerns of using AI for assessment, and whether it would affect workers’ registrations as social workers.</p><p>“So we kind of approached it very much in a sort of we'll have a volunteer from every team, and they are our champions for that team, knowing that social workers work very much in teams,” Ropstad says, adding that workers are now reassured that their jobs aren’t at risk because of Copilot.</p><p>All three councils are collaborating and continue to work with Microsoft on furthering their use of Copilot and getting the most out of the tool.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/how-welsh-councils-are-improving-services-with-microsoft-copilot</link>
                                                                            <description>
                            <![CDATA[ AI use has reduced staff toil, improved customer service, and increased team collaboration at three Welsh councils ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">EVJJzqeyKG4oqq3fXkR6EB</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/TNMEsRRa95BCbgz4PpLTjk-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 15 May 2026 07:00:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:30 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ rory.bathgate@futurenet.com (Rory Bathgate) ]]></author>                    <dc:creator><![CDATA[ Rory Bathgate ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Vep5JogbPhduK7R6CUWAm6-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Rory Bathgate was the Features and Multimedia Editor at ITPro until 2026, overseeing all in-depth content and case studies. Throughout his time at ITPro, Rory  charted the rise in popularity of generative AI and specifically companies such as Microsoft, OpenAI, and Google. &lt;/p&gt;&lt;p&gt;Alongside this, he delved into increasing calls for ethical and responsible AI as global legislators circle the technology, as well as the latest in mobile networking technology, from 5G mmWave to the 3G sunset and how it will affect businesses.&lt;/p&gt;&lt;p&gt;He  provided coverage from high-profile tech conferences such as Dell Technologies World, SuiteWorld, and VMware Explore Europe. His on-the-ground coverage  included live blogs, extensive daily coverage of the most significant announcements, analysis pieces, and podcast.&lt;/p&gt;&lt;p&gt;Rory was the co-host of the ITPro Podcast alongside Jane McCallion, where he swapped a keyboard for a microphone to discuss the latest learnings in tech.&lt;/p&gt;&lt;p&gt;Rory has an MA in Eighteenth-Century Studies from King’s College London, as well as a BA in English and American Literature from the University of Kent. He joined ITPro in 2022 as a graduate, after four years in student journalism.&lt;/p&gt;&lt;p&gt;In his free time, Rory enjoys photography and video editing, and can often be found at the cinema or reading a good science fiction paperback.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/TNMEsRRa95BCbgz4PpLTjk-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A scenic view of Dryslwyn Castle in Carmarthenshire, Wales.]]></media:description>                                                            <media:text><![CDATA[A scenic view of Dryslwyn Castle in Carmarthenshire, Wales.]]></media:text>
                                <media:title type="plain"><![CDATA[A scenic view of Dryslwyn Castle in Carmarthenshire, Wales.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/TNMEsRRa95BCbgz4PpLTjk-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Microsoft 365 Copilot is becoming widespread, as the go-to <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> service at firms already invested in the M365 productivity suite. The AI tool is the first port of call many organizations will have with generative AI and in the right hands, can drive meaningful increases in productivity.</p><p>Some standout examples of benefits can be found in the public sector, where pioneering councils are already using Copilot to the benefit of their staff.</p><p><em>ITPro</em> spoke to representatives to three such councils in Wales, who have led the charge on AI adoption in their respective roles: Mari Ropstad, head of service at Rhondda Cynon Taf County Borough Council; Ness Young, director of Corporate Service, City &amp; County of Swansea; and Georgia Sweet, digital transformation project officer, Carmarthenshire County Council.</p><h2 id="concrete-benefits">Concrete benefits</h2><p>Microsoft figures show that across the councils are already saving time and improving processes using Copilot.</p><p>Swansea County Council estimates that it saved 5,400 staff hours in just four weeks, as a direct effect of AI usage.</p><p>In Rhondda Cynon Taf County Borough Council, leaders have noted marked improvements in core tasks such as assessments, which Copilot has allowed to be created almost four times faster. After adopting the tool, 75% of staff say they feel more positive about their role, with 86% also reporting the quality of their work has improved.</p><p>At Carmarthenshire County Council, Sweet explains her team has developed a Copilot agent that responds to customer enquiries with FAQs drawn from the website. The deployment has helped the council to meet its ten-day target for complaints.</p><p>In addition to improving the customer-facing experience for Carmarthenshire County Council, Copilot has also brought benefits for workers inside the organization.</p><p>“We've seen a lot of feedback from <a href="https://www.itpro.com/business/careers-and-training/how-can-leaders-support-neurodiversity-in-tech"><u>neurodiverse</u></a> colleagues, particularly those with ADHD and dyslexia, where they're finding that Copilot is transforming the way that they work,” Sweet says.</p><p>“So they're finding it easy to process information, it's really helping with their written communication and the tools like rewrite or summarizing large documents, that's really helping them and improving their experience at work.”</p><p>Sweet adds that Copilot can be used without technical experience, with options to take it further depending on the user’s skill level.</p><h2 id="teething-difficulties">Teething difficulties</h2><p>Despite Copilot’s relative ease of access, users do need a basic understanding of Microsoft and AI systems to make the most of the tool – and for some staff, this was an initial stumbling block. </p><p>Young says that adoption of Copilot has revealed that some staff are less digitally literate than expected, necessitating additional time for training.</p><p>“We've got about 3,500 people using Copilot in Swansea, and some of the feedback that we've had from those providing training and engagement is that digital literacy can be an obstacle to initial use,” Young tells <em>ITPro</em>.</p><p>“Not necessarily just in a technical sense, but also in terms of confidence. And so we have invested heavily, we've got a support partner working with us to train people and we also have an engagement officer who provides one to one support particularly for those who have full licenses.”</p><p>Ropstad’s experience has been similar, with her team having found that users need “concrete use cases” for AI tools beyond basic training, including suggested prompts for generating forms or summarizing notes.</p><p>“So we use a lot of our own practitioners to teach other practitioners,” Ropstad explains. “And for us, it's always about trying to bring people into this middle band. You know, we are a council we can't afford for quality to be haphazard, it has to be a standard of quality. And so in order to get everybody to a middle band of Copilot use, we have to be able to bring everybody up to that level.”</p><p>All three councils benefited from first providing Copilot to a small group of ‘hero’ users, those most enthusiastic about AI and with the greatest technical experience, who helped to identify the most promising avenues for the technology and down the line to share their expertise with colleagues.</p><p>At Swansea Council, Young says 100 licenses were initially assigned, with 75 given to workers on pilot projects identified through discussions with Microsoft and other Welsh councils. The remaining 25 were assigned to the council’s senior leadership team, to make it a “leadership opportunity and really drive it from the top of the organization”, Young adds.</p><p>Carmarthenshire County Council started with a similarly small group, who later shared their learnings with the wider organization. Internally, employees are now sharing prompts via Teams, as well as ideas for AI use.</p><p>“I think people are so keen to share their learnings and if they found an amazing prompt, or if they've got a new process they're using Copilot for, and I think that's quite rare with new technology, to have that kind of buzz and excitement,” Sweet says.</p><p>“It’s as though the product sort of sells itself,” Young adds.</p><p>At Rhondda Cynon Taf County Borough Council, around 350 staff are now using Microsoft 365 Copilot. Initial concerns at the council centered around the ethical concerns of using AI for assessment, and whether it would affect workers’ registrations as social workers.</p><p>“So we kind of approached it very much in a sort of we'll have a volunteer from every team, and they are our champions for that team, knowing that social workers work very much in teams,” Ropstad says, adding that workers are now reassured that their jobs aren’t at risk because of Copilot.</p><p>All three councils are collaborating and continue to work with Microsoft on furthering their use of Copilot and getting the most out of the tool.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ OpenAI ramps up enterprise AI push with new consultancy launch ]]></title>
                                                                                                <dc:content><![CDATA[ <p>OpenAI is making a play for enterprise business by setting up its own consultancy to help organizations roll out AI. </p><p>The new OpenAI Deployment Company will support engineers specializing in frontier AI deployment – known as Forward Deployed Engineers (FDEs) – who can work directly with organizations integrating the technology to speed up the process.</p><p>"These FDEs will work closely with business leaders, operators, and frontline teams to identify where AI can make the biggest impact, redesign organizational infrastructure and critical workflows around it, and turn those gains into durable systems," OpenAI said in a <a href="https://openai.com/index/openai-launches-the-deployment-company/" target="_blank"><u>blog post</u></a>. </p><p>The move by OpenAI comes as it targets deeper ties with business customers. Last month, the AI developer revealed that enterprise business now makes up  40% of its revenue, and would reach half by the end of this year – key for the company as it <a href="https://www.itpro.com/business/business-strategy/openai-cfo-teases-new-money-making-plans-as-ai-giant-targets-practical-adoption-gains-in-2026-and-beyond"><u>chases new sources of income</u></a>. </p><p>Denise Dresser, chief revenue officer at OpenAI, said that AI is increasingly capable of actually doing meaningful work. </p><p>"The challenge now is helping companies integrate these systems into the infrastructure and workflows that power their businesses," she said in a statement. </p><p>"DeployCo is designed to help organizations bridge that gap and turn AI capability into real operational impact."</p><h2 id="how-openai-s-new-consultancy-will-work">How OpenAI’s new consultancy will work</h2><p>According to OpenAI, the consultancy will start with a "focused diagnostic" to find out where AI can bring the most value, selecting priority workflows to rethink with input from leadership and employees. </p><p>Speaking to <em>CNBC </em>ahead of the launch, Dresser said the ultimate aim here is to offer a helping hand to enterprises and streamline adoption – an issue many have struggled with over the last three years. </p><p>"Think about the complex workflows, about how you actually build a product service, a product market, a product, and this structure of this company is going to allow us to do that at speed and scale," she told <a href="https://www.cnbc.com/2026/05/11/open-ai-dresser-enterprise-business.html" target="_blank"><u><em>CNBC</em></u></a>. </p><p>FDEs will be embedded within companies to understand how the organization works, helping to design, build, test and deploy AI in their workflows in a reliable way that works with existing business processes. </p><p>"Forward-deployed engineers can sit with an organization, sit with their users, understand the workflow, and then help them take that capability from their back-office applications, connecting it to the model, and then really building intelligence in terms of each of the workflow," Dresser told <em>CNBC</em>.</p><h2 id="acquisition-of-tomoro">Acquisition of Tomoro</h2><p>As part of the launch, OpenAI has acquired <a href="https://www.itpro.com/technology/artificial-intelligence/agentic-ai-tools-gartner-agent-washing">AI consultancy</a> and engineering firm Tomoro, which includes 150 FDEs and has previously worked for Tesco, Virgin Atlantic, and Supercell. </p><p>The new company will partner with 19 additional investment firms, consultancies and system integrators, including TPG, Advent, Bain Capital, and Brookfield, with £4 billion in investment from Capgemini, McKinsey, and more.</p><p>OpenAI holds a majority stake, in order to give customers a "unified experience" whether they work directly with OpenAI or the deployment company. </p><p>The OpenAI Deployment Company will operate as a standalone business unit to help meet corporate pace and requirements, but will remain an extension of OpenAI to ensure customers have access to the latest research – and know what's coming next. </p><p>"The OpenAI Deployment Company FDEs will be able to build for where OpenAI’s frontier capabilities are headed, giving customers systems designed to improve as new models, tools, and deployment patterns come online," the blog post added. </p><p>"Customers can move faster from day one, spend capital on durable systems, and stay ahead of competitors by building around the capabilities that are coming next."</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/business/business-strategy/openai-ramps-up-enterprise-ai-push-with-new-consultancy-launch</link>
                                                                            <description>
                            <![CDATA[ The launch of the OpenAI Deployment Company comes with £4bn investment to help speed up enterprise Ai adoption ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">QN5Hx9tKGWxtydYmfc8quC</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/yzNg5FNmrTNkz3EiQpZWjP-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 13 May 2026 07:30:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 15:00:15 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/yzNg5FNmrTNkz3EiQpZWjP-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Close-up shot of the OpenAI logo pictured on a smartphone screen.]]></media:description>                                                            <media:text><![CDATA[Close-up shot of the OpenAI logo pictured on a smartphone screen.]]></media:text>
                                <media:title type="plain"><![CDATA[Close-up shot of the OpenAI logo pictured on a smartphone screen.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/yzNg5FNmrTNkz3EiQpZWjP-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>OpenAI is making a play for enterprise business by setting up its own consultancy to help organizations roll out AI. </p><p>The new OpenAI Deployment Company will support engineers specializing in frontier AI deployment – known as Forward Deployed Engineers (FDEs) – who can work directly with organizations integrating the technology to speed up the process.</p><p>"These FDEs will work closely with business leaders, operators, and frontline teams to identify where AI can make the biggest impact, redesign organizational infrastructure and critical workflows around it, and turn those gains into durable systems," OpenAI said in a <a href="https://openai.com/index/openai-launches-the-deployment-company/" target="_blank"><u>blog post</u></a>. </p><p>The move by OpenAI comes as it targets deeper ties with business customers. Last month, the AI developer revealed that enterprise business now makes up  40% of its revenue, and would reach half by the end of this year – key for the company as it <a href="https://www.itpro.com/business/business-strategy/openai-cfo-teases-new-money-making-plans-as-ai-giant-targets-practical-adoption-gains-in-2026-and-beyond"><u>chases new sources of income</u></a>. </p><p>Denise Dresser, chief revenue officer at OpenAI, said that AI is increasingly capable of actually doing meaningful work. </p><p>"The challenge now is helping companies integrate these systems into the infrastructure and workflows that power their businesses," she said in a statement. </p><p>"DeployCo is designed to help organizations bridge that gap and turn AI capability into real operational impact."</p><h2 id="how-openai-s-new-consultancy-will-work">How OpenAI’s new consultancy will work</h2><p>According to OpenAI, the consultancy will start with a "focused diagnostic" to find out where AI can bring the most value, selecting priority workflows to rethink with input from leadership and employees. </p><p>Speaking to <em>CNBC </em>ahead of the launch, Dresser said the ultimate aim here is to offer a helping hand to enterprises and streamline adoption – an issue many have struggled with over the last three years. </p><p>"Think about the complex workflows, about how you actually build a product service, a product market, a product, and this structure of this company is going to allow us to do that at speed and scale," she told <a href="https://www.cnbc.com/2026/05/11/open-ai-dresser-enterprise-business.html" target="_blank"><u><em>CNBC</em></u></a>. </p><p>FDEs will be embedded within companies to understand how the organization works, helping to design, build, test and deploy AI in their workflows in a reliable way that works with existing business processes. </p><p>"Forward-deployed engineers can sit with an organization, sit with their users, understand the workflow, and then help them take that capability from their back-office applications, connecting it to the model, and then really building intelligence in terms of each of the workflow," Dresser told <em>CNBC</em>.</p><h2 id="acquisition-of-tomoro">Acquisition of Tomoro</h2><p>As part of the launch, OpenAI has acquired <a href="https://www.itpro.com/technology/artificial-intelligence/agentic-ai-tools-gartner-agent-washing">AI consultancy</a> and engineering firm Tomoro, which includes 150 FDEs and has previously worked for Tesco, Virgin Atlantic, and Supercell. </p><p>The new company will partner with 19 additional investment firms, consultancies and system integrators, including TPG, Advent, Bain Capital, and Brookfield, with £4 billion in investment from Capgemini, McKinsey, and more.</p><p>OpenAI holds a majority stake, in order to give customers a "unified experience" whether they work directly with OpenAI or the deployment company. </p><p>The OpenAI Deployment Company will operate as a standalone business unit to help meet corporate pace and requirements, but will remain an extension of OpenAI to ensure customers have access to the latest research – and know what's coming next. </p><p>"The OpenAI Deployment Company FDEs will be able to build for where OpenAI’s frontier capabilities are headed, giving customers systems designed to improve as new models, tools, and deployment patterns come online," the blog post added. </p><p>"Customers can move faster from day one, spend capital on durable systems, and stay ahead of competitors by building around the capabilities that are coming next."</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Inforcer launches new Copilot Manager to help MSPs scale AI services ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Inforcer has cut the ribbon on a new Copilot Manager feature aimed at helping <a href="https://www.itpro.com/business/have-we-seen-the-end-of-the-true-msp">managed service providers (MSPs)</a> develop and scale secure AI services.</p><p>Available to demo now, the tool is designed to provide greater visibility into how Microsoft Copilot is being used across customer environments, while also enabling them to monitor and manage ‘shadow AI’ use – where employees use <a href="https://www.itpro.com/technology/artificial-intelligence/in-from-the-shadows-ais-uncontrolled-growth">unauthorized AI tools</a> as part of their work.</p><p>As SMB demand for secure and profitable AI services continues to grow, Inforcer said Copilot Manager allows partners to track adoption trends, identify power users and opportunities for further licensing, while highlighting return on investment (ROI) for customers.</p><p>The tool also provides insight into which shadow AI applications are being used, how frequently, as well as just how much data is being shared, helping MSPs to identify potential security risks and compliance gaps.</p><p>“MSPs need Copilot support now,” commented Jamie Daum, CEO at Inforcer. “Time is of the essence for MSPs to implement innovative solutions that capitalise on emerging AI opportunities, ensuring they stay competitive in a rapidly evolving market.”</p><h2 id="addressing-the-shadow-ai-challenge">Addressing the shadow AI challenge</h2><p>The new tool comes in response to the rise in SMB employees using unauthorized and open-source AI tools as part of their work, increasing the potential threat of data leakage.</p><p>As part of its wider Microsoft 365 management platform, Inforcer said Copilot Manager will enable MSPs to turn this growing challenge into a commercial opportunity by helping customers identify risks, enforce governance policies, and implement more secure AI usage practices.</p><p>The offering allows partners to assess shadow AI usage across multiple tenants, flag high-risk applications, as well as recommend data protection and compliance measures as part of their managed services. MSPs can also leverage the tool to upsell additional services such as governance projects and security enhancements.</p><p>Additionally, the feature complements Inforcer’s existing Copilot Readiness Assessment tool, which helps MSPs evaluate the maturity of customer environments prior to deploying AI services.</p><p>The firm said these combined capabilities will help MSPs support clients at every stage of their <a href="https://www.itpro.com/security/safe-ai-adoption-rests-on-cybersecurity-professionals-says-rsac-chairman">AI adoption</a> – from initial readiness through to ongoing optimization and governance.</p><h2 id="bridging-the-copilot-visibility-gap">Bridging the Copilot visibility gap</h2><p>Inforcer’s Copilot Manager tool has already been trialled in beta by more than 200 MSPs around the world. </p><p>Darron Millar, solution director at UK-based digital transformation specialist BCN Group, said the offering’s Copilot and shadow AI usage insights enable the company to balance value creation with real-world risk awareness.</p><p>“This data complements our managed security and productivity services and supports our initiatives around Copilot readiness and safe adoption,” he explained.</p><p>“Having trusted, tenant-level insight means we can move beyond assumptions and provide customers with informed, actionable guidance on where to invest, where to control risk, and how to maximise return from Copilot.”</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/inforcer-launches-new-copilot-manager-to-help-msps-scale-ai-services</link>
                                                                            <description>
                            <![CDATA[ The tool aims to give greater visibility into Copilot usage and shadow AI risks across customer environments ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">P7p4b6kBTiSKXhJWZWVJC8</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/7K6HfpqyunGBX9PzbBirqL-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 29 Apr 2026 14:53:28 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:07 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ itpro@futurenet.com (Daniel Todd) ]]></author>                    <dc:creator><![CDATA[ Daniel Todd ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/SRyC34qeLpNDj3dJtsVDhT-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/7K6HfpqyunGBX9PzbBirqL-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[AI productivity concept image showing a small human standing on the fingertip of a large robotic hand while interacting with digital interfaces.]]></media:description>                                                            <media:text><![CDATA[AI productivity concept image showing a small human standing on the fingertip of a large robotic hand while interacting with digital interfaces.]]></media:text>
                                <media:title type="plain"><![CDATA[AI productivity concept image showing a small human standing on the fingertip of a large robotic hand while interacting with digital interfaces.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/7K6HfpqyunGBX9PzbBirqL-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Inforcer has cut the ribbon on a new Copilot Manager feature aimed at helping <a href="https://www.itpro.com/business/have-we-seen-the-end-of-the-true-msp">managed service providers (MSPs)</a> develop and scale secure AI services.</p><p>Available to demo now, the tool is designed to provide greater visibility into how Microsoft Copilot is being used across customer environments, while also enabling them to monitor and manage ‘shadow AI’ use – where employees use <a href="https://www.itpro.com/technology/artificial-intelligence/in-from-the-shadows-ais-uncontrolled-growth">unauthorized AI tools</a> as part of their work.</p><p>As SMB demand for secure and profitable AI services continues to grow, Inforcer said Copilot Manager allows partners to track adoption trends, identify power users and opportunities for further licensing, while highlighting return on investment (ROI) for customers.</p><p>The tool also provides insight into which shadow AI applications are being used, how frequently, as well as just how much data is being shared, helping MSPs to identify potential security risks and compliance gaps.</p><p>“MSPs need Copilot support now,” commented Jamie Daum, CEO at Inforcer. “Time is of the essence for MSPs to implement innovative solutions that capitalise on emerging AI opportunities, ensuring they stay competitive in a rapidly evolving market.”</p><h2 id="addressing-the-shadow-ai-challenge">Addressing the shadow AI challenge</h2><p>The new tool comes in response to the rise in SMB employees using unauthorized and open-source AI tools as part of their work, increasing the potential threat of data leakage.</p><p>As part of its wider Microsoft 365 management platform, Inforcer said Copilot Manager will enable MSPs to turn this growing challenge into a commercial opportunity by helping customers identify risks, enforce governance policies, and implement more secure AI usage practices.</p><p>The offering allows partners to assess shadow AI usage across multiple tenants, flag high-risk applications, as well as recommend data protection and compliance measures as part of their managed services. MSPs can also leverage the tool to upsell additional services such as governance projects and security enhancements.</p><p>Additionally, the feature complements Inforcer’s existing Copilot Readiness Assessment tool, which helps MSPs evaluate the maturity of customer environments prior to deploying AI services.</p><p>The firm said these combined capabilities will help MSPs support clients at every stage of their <a href="https://www.itpro.com/security/safe-ai-adoption-rests-on-cybersecurity-professionals-says-rsac-chairman">AI adoption</a> – from initial readiness through to ongoing optimization and governance.</p><h2 id="bridging-the-copilot-visibility-gap">Bridging the Copilot visibility gap</h2><p>Inforcer’s Copilot Manager tool has already been trialled in beta by more than 200 MSPs around the world. </p><p>Darron Millar, solution director at UK-based digital transformation specialist BCN Group, said the offering’s Copilot and shadow AI usage insights enable the company to balance value creation with real-world risk awareness.</p><p>“This data complements our managed security and productivity services and supports our initiatives around Copilot readiness and safe adoption,” he explained.</p><p>“Having trusted, tenant-level insight means we can move beyond assumptions and provide customers with informed, actionable guidance on where to invest, where to control risk, and how to maximise return from Copilot.”</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Accenture to roll-out Copilot to 700,000+ staff ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Microsoft has revealed Accenture will equip its entire workforce with <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-copilot-review-ai-baked-into-your-apps">Copilot </a>in a deal that marks its biggest to date. </p><p>The professional services company ranks among one of the early adopters of Microsoft’s flagship AI service, having slowly rolled out its use across the consultancy over the last three years. </p><p>The roll-out initially saw just a few hundred select employees and senior leaders given access to Copilot, the company noted in a blog post. </p><p>That was then scaled up to around 20,000 users, with Accenture focusing on <a href="https://www.itpro.com/technology/artificial-intelligence/data-strategy-ai-adoption-failures">data strategy</a>, governance, and access controls, as well as understanding how staff actually made use of the tools. </p><p>“We were fine-tuning our adoption strategy and developing a blueprint for how it would be used in daily work,” said Tony Leraris, Accenture’s <a href="https://www.itpro.com/tag/chief-information-officer">chief information officer (CIO)</a>.</p><p>“You can’t take a one-size-fits-all message into adoption. We really had to demonstrate to certain people, especially leaders, how to use the tool and what the value would be specifically for them,” he added. </p><p>As the company scaled up to 200,000 by the end of last year, it focused on change management, including making use of Viva Engage, a social networking app built into Microsoft 365. </p><p>Now, Accenture aims to cover the vast majority of its workforce, some 743,000 people across 120 countries – equivalent to the population of Denver, Colorado. </p><h2 id="accenture-touts-copilot-gains">Accenture touts Copilot gains</h2><p>In a company survey last year, 97% of Accenture employees said they completed routine tasks 15 times faster thanks to the AI tool, with 53% reporting significant productivity and efficiency improvements. </p><p>Accenture said one set of 200,000 licenses had monthly usage of 89%, with a survey showing 84% of staff would "deeply miss" the tool if it was removed. </p><p>"If <a href="https://www.itpro.com/desktop-software/19337/office-365-review">Microsoft 365</a> Copilot weren’t delivering real value, our people simply wouldn’t be using it – our high adoption rate is what shows us that there is value," Leraris said. “That’s what led us to continue deploying Copilot to more people."</p><p>The Accenture deal marks a significant milestone for Microsoft as it ramps up efforts to roll-out <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a> to a wider audience –<strong> </strong>a strategy that some reports last year suggested was floundering. </p><p>Reports from <a href="https://www.theinformation.com/articles/microsoft-lowers-ai-software-sales-quotas-customers-resist-newer-products" target="_blank"><u><em>The Information</em></u></a> in December noted that the tech giant lowered sales growth targets for agentic AI products after salespeople missed quotas.</p><p>Microsoft has also announced a number of deals aimed at attracting wider appeal for core product ranges, including agreements to integrate Anthropic’s Claude AI within Copilot, and most recently with Claude Cowork features. </p><p>In March, the tech giant <a href="https://www.itpro.com/technology/artificial-intelligence/anthropics-claude-cowork-tool-is-coming-to-microsoft-copilot"><u>unveiled Copilot Cowork</u></a>, a fusion of Anthropic’s agentic AI offering with its own in-house chatbot. </p><h2 id="not-everyone-is-sold-on-ai">Not everyone is sold on AI</h2><p>While Accenture has recorded marked success with Copilot, not everyone has seen similarly positive results. A <a href="https://www.itpro.com/business/business-strategy/ai-productivity-business-nber-study-white-collar-work"><u>survey by the National Bureau of Economic Research</u></a>, for example, revealed AI's impact remains limited in terms of productivity. </p><p>Notably, Accenture’s own recent <a href="https://www.accenture.com/gb-en/insights/data-ai/generating-impact" target="_blank"><u><em>Generating Impact</em></u></a><em> </em>report highlighted a “widening gap” in terms of AI’s effectiveness in enterprise settings. </p><p>Many staff are using the technology for basic tasks, but its broader impact on organizational change and performance is often limited. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/accenture-has-been-trialling-microsoft-copilot-since-2023-now-its-rolling-out-the-ai-tool-to-all-743-000-staff</link>
                                                                            <description>
                            <![CDATA[ Accenture will roll out Microsoft Copilot to nearly three quarters of a million employees after years of testing ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">Znmd8gAw4aWpyXoFeeF5Eo</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/efchWqcCFapy7tbZiwPZ3F-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 28 Apr 2026 15:17:34 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:41 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/efchWqcCFapy7tbZiwPZ3F-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Illuminated Accenture logo pictured at CES 2026 in Las Vegas, with silhouetted man walking past in low light.]]></media:description>                                                            <media:text><![CDATA[Illuminated Accenture logo pictured at CES 2026 in Las Vegas, with silhouetted man walking past in low light.]]></media:text>
                                <media:title type="plain"><![CDATA[Illuminated Accenture logo pictured at CES 2026 in Las Vegas, with silhouetted man walking past in low light.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/efchWqcCFapy7tbZiwPZ3F-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Microsoft has revealed Accenture will equip its entire workforce with <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-copilot-review-ai-baked-into-your-apps">Copilot </a>in a deal that marks its biggest to date. </p><p>The professional services company ranks among one of the early adopters of Microsoft’s flagship AI service, having slowly rolled out its use across the consultancy over the last three years. </p><p>The roll-out initially saw just a few hundred select employees and senior leaders given access to Copilot, the company noted in a blog post. </p><p>That was then scaled up to around 20,000 users, with Accenture focusing on <a href="https://www.itpro.com/technology/artificial-intelligence/data-strategy-ai-adoption-failures">data strategy</a>, governance, and access controls, as well as understanding how staff actually made use of the tools. </p><p>“We were fine-tuning our adoption strategy and developing a blueprint for how it would be used in daily work,” said Tony Leraris, Accenture’s <a href="https://www.itpro.com/tag/chief-information-officer">chief information officer (CIO)</a>.</p><p>“You can’t take a one-size-fits-all message into adoption. We really had to demonstrate to certain people, especially leaders, how to use the tool and what the value would be specifically for them,” he added. </p><p>As the company scaled up to 200,000 by the end of last year, it focused on change management, including making use of Viva Engage, a social networking app built into Microsoft 365. </p><p>Now, Accenture aims to cover the vast majority of its workforce, some 743,000 people across 120 countries – equivalent to the population of Denver, Colorado. </p><h2 id="accenture-touts-copilot-gains">Accenture touts Copilot gains</h2><p>In a company survey last year, 97% of Accenture employees said they completed routine tasks 15 times faster thanks to the AI tool, with 53% reporting significant productivity and efficiency improvements. </p><p>Accenture said one set of 200,000 licenses had monthly usage of 89%, with a survey showing 84% of staff would "deeply miss" the tool if it was removed. </p><p>"If <a href="https://www.itpro.com/desktop-software/19337/office-365-review">Microsoft 365</a> Copilot weren’t delivering real value, our people simply wouldn’t be using it – our high adoption rate is what shows us that there is value," Leraris said. “That’s what led us to continue deploying Copilot to more people."</p><p>The Accenture deal marks a significant milestone for Microsoft as it ramps up efforts to roll-out <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a> to a wider audience –<strong> </strong>a strategy that some reports last year suggested was floundering. </p><p>Reports from <a href="https://www.theinformation.com/articles/microsoft-lowers-ai-software-sales-quotas-customers-resist-newer-products" target="_blank"><u><em>The Information</em></u></a> in December noted that the tech giant lowered sales growth targets for agentic AI products after salespeople missed quotas.</p><p>Microsoft has also announced a number of deals aimed at attracting wider appeal for core product ranges, including agreements to integrate Anthropic’s Claude AI within Copilot, and most recently with Claude Cowork features. </p><p>In March, the tech giant <a href="https://www.itpro.com/technology/artificial-intelligence/anthropics-claude-cowork-tool-is-coming-to-microsoft-copilot"><u>unveiled Copilot Cowork</u></a>, a fusion of Anthropic’s agentic AI offering with its own in-house chatbot. </p><h2 id="not-everyone-is-sold-on-ai">Not everyone is sold on AI</h2><p>While Accenture has recorded marked success with Copilot, not everyone has seen similarly positive results. A <a href="https://www.itpro.com/business/business-strategy/ai-productivity-business-nber-study-white-collar-work"><u>survey by the National Bureau of Economic Research</u></a>, for example, revealed AI's impact remains limited in terms of productivity. </p><p>Notably, Accenture’s own recent <a href="https://www.accenture.com/gb-en/insights/data-ai/generating-impact" target="_blank"><u><em>Generating Impact</em></u></a><em> </em>report highlighted a “widening gap” in terms of AI’s effectiveness in enterprise settings. </p><p>Many staff are using the technology for basic tasks, but its broader impact on organizational change and performance is often limited. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ The first hurdle is the hardest in generative AI adoption – and businesses keep falling ]]></title>
                                                                                                <dc:content><![CDATA[ <p>AI adoption is ramping up at a rapid pace globally, of that there’s no doubt. But there’s more to successful adoption than just buying shiny new tools and finding yourself in a productivity paradise. </p><p>While that might seem like stating the obvious, it’s still a problem a significant portion of enterprises find themselves facing – and there’s little sign that it will improve any time soon. </p><p>Research from Amazon Web Services (AWS), released to coincide with its annual AWS Summit London this week, showed progress on the adoption front compared to last year. </p><p>The study, which focuses primarily on UK-based firms, has some positive takeaways, with British enterprises outpacing adoption compared to their European counterparts. </p><p>Clearcut gains with the technology are also being made, according to AWS, with workers now basking in the productivity and efficiency gains promised by it and other big tech providers. </p><p>Speaking during the keynote address at the summit this week, Alison Kay, AWS’ VP and managing director for UK&amp;I, waxed lyrical about the benefits of AI for customers and the company itself. </p><p>Developers are working at lightning speed delivering projects previously deemed insurmountable while customers such as Motorway are recording the same gains. </p><p>“We’re living through just an incredible moment in history, a breakthrough moment in technology advancement,” Kay told attendees. </p><p>“And what makes this moment so exciting is that the things that are being built, they almost defy belief. They almost feel like magic.”</p><p>AI might look and feel like magic, but IT leaders aren't living in a fantasy land – they're facing the harsh reality of how difficult AI implementation can be. They've got the trendy new tools and staff are loving them, yet they’re not seeing any significant progress on a business-wide level.</p><h2 id="ai-pilot-purgatory-persists">AI pilot purgatory persists</h2><p>AWS’ study found the actual use of AI in the enterprise is still very basic, with employees using the technology to summarize documents or answer basic questions using chatbots. </p><p>That’s so 2023, but at least they’re enjoying it. More than two-thirds (68%) of employees are reporting marked productivity and efficiency improvements. </p><p>AWS’ findings align closely with similar research from Accenture earlier this month, which pointed toward “localized” gains whereby individual workers are seeing gains, but the companies themselves are still struggling. </p><p>So what’s stopping enterprises from making the leap toward more “advanced” AI uses? The same old problems many businesses have with <a href="https://www.itpro.com/strategy/29899/three-reasons-why-digital-transformation-is-essential-for-business-growth">digital transformation</a> projects. </p><p>Skills and expertise, <a href="https://www.itpro.com/security/the-risks-of-legacy-technology-for-your-business">legacy systems</a>, and <a href="https://www.itpro.com/business/digital-transformation/disparate-data-silos-are-still-blocking-the-path-to-digital-transformation-success">outdated infrastructure</a> are stifling progress. </p><p>It’s been a common recurring theme with AI adoption over the last three and a half years and one that must have IT leaders bashing their heads off the wall over. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence/uk-firms-are-stuck-in-an-ai-pilot-purgatory-heres-how-they-can-drive-success"><u>Research from Google Cloud</u></a> in August 2024 warned British firms in particular were stuck in a “pilot purgatory”. </p><p>At that point, only two-in-five firms had pushed generative AI projects to production and were facing a “substantial hurdle” in scaling the technology internally - an issue the aforementioned Accenture study found is still a key problem. </p><p>The situation is even worse with agentic AI, with a January study from Dynatrace showing that half of agent adoption schemes are wallowing in the proof of concept (PoC) stage. </p><p>Phil Le-Brun, executive in residence at AWS, told <em>ITPro </em>that a “lack of imagination” is often a key factor in this inability to scale AI adoption and reach more advanced uses. </p><iframe allow="" height="200px" width="100%" id="" style="" class="position-center" data-lazy-priority="high" data-lazy-src="https://player.captivate.fm/episode/7b2774d8-c5b2-4912-a6aa-6dc1788dea4d/" src="https://player.captivate.fm/episode/7b2774d8-c5b2-4912-a6aa-6dc1788dea4d/"></iframe><p>An increasing volume of research also points toward the frantic nature of the <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> boom as a key factor behind failed adoption efforts. </p><p><a href="https://www.itpro.com/business/business-strategy/enterprises-are-paralyzed-by-a-lack-of-understanding-with-ai-adoption-and-theres-one-key-factor-that-decides-success"><u>Research from Forrester</u></a> in early April suggested that a “lack of understanding” when it comes to AI, combined with the rabid appetite to invest in the technology, has resulted in a slap-dash approach to AI innovation. </p><p>Executives have been caught up in the hype, and all they can see are dollar signs and the marketing materials promising huge productivity improvements. </p><p>But Forrester noted those who take a more conservative, “deliberate” approach to AI adoption have recorded far better results – and a key factor here is focusing on creating value for customers rather than productivity and cost reductions. </p><p>Speaking to <em>ITPro</em>, Le-Brun touched on this customer-centric approach with AI, noting that some of the more agile companies it engages with, such as startups, are making great strides in this regard. </p><p>Simply put, keep an eye on the productivity and cost benefits, but sharpen your focus on actually delivering for customers. </p><p>“The application of AI to fundamentally rethink how you’re delivering a product or service to the customer would be an example of advanced AI,” he said. “That’s where we see startups doing some amazing things.”</p><p>“They’re willing to really dive into what it is the customer wants, and how can we apply this new technology to change, rather than for large enterprises - it’s much harder because they have a lot of established processes.”</p><h2 id="a-long-road-ahead">A long road ahead</h2><p>The current pace of progress in pushing from pilots and basic use-cases toward more advanced AI uses highlights the sorry state of affairs for many businesses - and there’s still a long road ahead. A very long one, if AWS findings are to be taken at face value. </p><p>Just 24% of firms have reached an advanced AI phase in 2026, according to AWS, marking just a single percentage increase compared to last year. If things continue at this pace of progress, it’ll be 2102 before all AI adopters in the UK reach an advanced stage. </p><p>As Le-Brun told <em>ITPro</em>, businesses will be in a good position “by the end of the century”. </p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/business/business-strategy/the-first-hurdle-is-the-hardest-in-generative-ai-adoption-and-businesses-keep-falling</link>
                                                                            <description>
                            <![CDATA[ AWS’ UK chief said AI advances “feel like magic” at its recent London summit, but many firms are facing the reality of sluggish gains ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">432hD26emDP4sw7nZX5jda</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/5U6eXUHt9vPgHT7z5pghjg-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 23 Apr 2026 17:30:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 15:00:19 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/5U6eXUHt9vPgHT7z5pghjg-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Man jumping over a hurdle with three fallen hurdles behind him.]]></media:description>                                                            <media:text><![CDATA[Man jumping over a hurdle with three fallen hurdles behind him.]]></media:text>
                                <media:title type="plain"><![CDATA[Man jumping over a hurdle with three fallen hurdles behind him.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/5U6eXUHt9vPgHT7z5pghjg-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>AI adoption is ramping up at a rapid pace globally, of that there’s no doubt. But there’s more to successful adoption than just buying shiny new tools and finding yourself in a productivity paradise. </p><p>While that might seem like stating the obvious, it’s still a problem a significant portion of enterprises find themselves facing – and there’s little sign that it will improve any time soon. </p><p>Research from Amazon Web Services (AWS), released to coincide with its annual AWS Summit London this week, showed progress on the adoption front compared to last year. </p><p>The study, which focuses primarily on UK-based firms, has some positive takeaways, with British enterprises outpacing adoption compared to their European counterparts. </p><p>Clearcut gains with the technology are also being made, according to AWS, with workers now basking in the productivity and efficiency gains promised by it and other big tech providers. </p><p>Speaking during the keynote address at the summit this week, Alison Kay, AWS’ VP and managing director for UK&amp;I, waxed lyrical about the benefits of AI for customers and the company itself. </p><p>Developers are working at lightning speed delivering projects previously deemed insurmountable while customers such as Motorway are recording the same gains. </p><p>“We’re living through just an incredible moment in history, a breakthrough moment in technology advancement,” Kay told attendees. </p><p>“And what makes this moment so exciting is that the things that are being built, they almost defy belief. They almost feel like magic.”</p><p>AI might look and feel like magic, but IT leaders aren't living in a fantasy land – they're facing the harsh reality of how difficult AI implementation can be. They've got the trendy new tools and staff are loving them, yet they’re not seeing any significant progress on a business-wide level.</p><h2 id="ai-pilot-purgatory-persists">AI pilot purgatory persists</h2><p>AWS’ study found the actual use of AI in the enterprise is still very basic, with employees using the technology to summarize documents or answer basic questions using chatbots. </p><p>That’s so 2023, but at least they’re enjoying it. More than two-thirds (68%) of employees are reporting marked productivity and efficiency improvements. </p><p>AWS’ findings align closely with similar research from Accenture earlier this month, which pointed toward “localized” gains whereby individual workers are seeing gains, but the companies themselves are still struggling. </p><p>So what’s stopping enterprises from making the leap toward more “advanced” AI uses? The same old problems many businesses have with <a href="https://www.itpro.com/strategy/29899/three-reasons-why-digital-transformation-is-essential-for-business-growth">digital transformation</a> projects. </p><p>Skills and expertise, <a href="https://www.itpro.com/security/the-risks-of-legacy-technology-for-your-business">legacy systems</a>, and <a href="https://www.itpro.com/business/digital-transformation/disparate-data-silos-are-still-blocking-the-path-to-digital-transformation-success">outdated infrastructure</a> are stifling progress. </p><p>It’s been a common recurring theme with AI adoption over the last three and a half years and one that must have IT leaders bashing their heads off the wall over. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence/uk-firms-are-stuck-in-an-ai-pilot-purgatory-heres-how-they-can-drive-success"><u>Research from Google Cloud</u></a> in August 2024 warned British firms in particular were stuck in a “pilot purgatory”. </p><p>At that point, only two-in-five firms had pushed generative AI projects to production and were facing a “substantial hurdle” in scaling the technology internally - an issue the aforementioned Accenture study found is still a key problem. </p><p>The situation is even worse with agentic AI, with a January study from Dynatrace showing that half of agent adoption schemes are wallowing in the proof of concept (PoC) stage. </p><p>Phil Le-Brun, executive in residence at AWS, told <em>ITPro </em>that a “lack of imagination” is often a key factor in this inability to scale AI adoption and reach more advanced uses. </p><iframe allow="" height="200px" width="100%" id="" style="" class="position-center" data-lazy-priority="high" data-lazy-src="https://player.captivate.fm/episode/7b2774d8-c5b2-4912-a6aa-6dc1788dea4d/" src="https://player.captivate.fm/episode/7b2774d8-c5b2-4912-a6aa-6dc1788dea4d/"></iframe><p>An increasing volume of research also points toward the frantic nature of the <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> boom as a key factor behind failed adoption efforts. </p><p><a href="https://www.itpro.com/business/business-strategy/enterprises-are-paralyzed-by-a-lack-of-understanding-with-ai-adoption-and-theres-one-key-factor-that-decides-success"><u>Research from Forrester</u></a> in early April suggested that a “lack of understanding” when it comes to AI, combined with the rabid appetite to invest in the technology, has resulted in a slap-dash approach to AI innovation. </p><p>Executives have been caught up in the hype, and all they can see are dollar signs and the marketing materials promising huge productivity improvements. </p><p>But Forrester noted those who take a more conservative, “deliberate” approach to AI adoption have recorded far better results – and a key factor here is focusing on creating value for customers rather than productivity and cost reductions. </p><p>Speaking to <em>ITPro</em>, Le-Brun touched on this customer-centric approach with AI, noting that some of the more agile companies it engages with, such as startups, are making great strides in this regard. </p><p>Simply put, keep an eye on the productivity and cost benefits, but sharpen your focus on actually delivering for customers. </p><p>“The application of AI to fundamentally rethink how you’re delivering a product or service to the customer would be an example of advanced AI,” he said. “That’s where we see startups doing some amazing things.”</p><p>“They’re willing to really dive into what it is the customer wants, and how can we apply this new technology to change, rather than for large enterprises - it’s much harder because they have a lot of established processes.”</p><h2 id="a-long-road-ahead">A long road ahead</h2><p>The current pace of progress in pushing from pilots and basic use-cases toward more advanced AI uses highlights the sorry state of affairs for many businesses - and there’s still a long road ahead. A very long one, if AWS findings are to be taken at face value. </p><p>Just 24% of firms have reached an advanced AI phase in 2026, according to AWS, marking just a single percentage increase compared to last year. If things continue at this pace of progress, it’ll be 2102 before all AI adopters in the UK reach an advanced stage. </p><p>As Le-Brun told <em>ITPro</em>, businesses will be in a good position “by the end of the century”. </p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ What is model collapse and why is it a risk for enterprise AI? ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Ever since <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> became mainstream, AI generated content has become increasingly common. From text, to audio, images, and video, <a href="https://www.itpro.com/technology/artificial-intelligence/what-is-synthetic-data"><u>synthetic data</u></a> has become easier than ever to produce.</p><p>A study by <a href="https://ahrefs.com/blog/what-percentage-of-new-content-is-ai-generated/"><u>Ahrefs</u></a>, published in 2025, analyzed 900,000 English-language web pages. About 2.5% of pages were purely AI-generated, and 71.7% of pages were a mix of human-written and AI-generated.</p><p>It's clear the percentage of AI-generated content, along with mixed content, is only going to increase. In some aspects, this is great news for enterprises, which can generate large, sanitized datasets that closely mimic their specific enterprise data without exposing actual private information to AI firms.</p><p>“Synthetic data shouldn’t be about replacing real data entirely but instead enhancing and extending it to build great AI models and agents,” says Dael Williamson, EMEA CTO at Databricks.</p><p>Ironically, however, the explosion of synthetic data has problematic implications for training future generations of AI models. Some firms, like Gartner, are already <a href="https://www.itpro.com/security/data-protection/fears-over-ai-model-collapse-are-fueling-a-shift-to-zero-trust-data-governance-strategies"><u>sounding the alarm</u></a> over how the massive volumes of AI-generated data coming down the pipeline. The concern is training on AI data could create a sort of algorithmic feedback loop, in which AI models become increasingly distanced from real data and more prone to <a href="https://www.itpro.com/technology/artificial-intelligence/ai-hallucinations-what-are-they"><u>hallucinations</u></a>. </p><p>A blend of intelligent AI-generated and human-written information is important. What if AI-generated information exhausts the percentage of real content, reaching its critical limit?</p><h2 id="what-is-model-collapse">What is model collapse?</h2><p>In a <a href="https://www.nature.com/articles/s41586-024-07566-y"><u>joint study</u></a>, Ilia Shumailov from the University of Oxford and Zakhar Shumaylov coined the term “model collapse”. </p><p>Model collapse is defined as a long-term statistical degenerative process in which an AI model gradually misperceives reality, bets on improbable events, and starts to generate low-quality or wrong outputs. An AI model begins to collapse when its training data gets contaminated by its output or other AI-generated information.</p><p>For example, model A trains on purely human-generated information. Model A now generates a large volume of synthetic data. A new model B now uses this synthetic data as a part of its training. Then, model C trains on synthetic data from models A and B… and so on. That’s how synthetic data ruins the training data distribution.</p><p>Anyone who ever tried to copy a VHS tape more than once has experienced generation loss, in which each reproduction of data produces a lesser copy filled with more and more artifacts. Model collapse could happen in much the same way, with AI developers forced to run more and more costly training runs to eke out performance increases in subsequent generations of frontier AI models.</p><p>Model collapse isn’t limited to LLMs: <a href="https://www.itpro.com/network-internet/29791/what-is-an-artificial-neural-network"><u>neural network</u></a> architectures such as variational encoders and probabilistic models such as mixture models are also susceptible.</p><p>A real world example can be found in OPT-125M, an <a href="https://www.itpro.com/technology/artificial-intelligence/three-open-source-large-language-models-you-can-use-today"><u>open large language model</u></a> created by Meta AI researchers. To test model collapse, the researchers trained OPT-125M on its own outputs, then prompted it to answer questions on historical architecture. In response, the model produced answers rabbits instead. In the near future, the nth generation of AI models would have to fight just such a collapse.</p><p>Williamson uses a real-world example to explain model collapse: </p><p>"Imagine a locksmith who keeps duplicating keys from the latest copy, not the original. Early on, the key still turns most locks.</p><p>“But after many rounds, the key may look right at a glance, yet it jams, scrapes, or spins uselessly. Copying copies collapses fidelity, but copying while anchored to the master preserves it”,<em> </em>Williamson adds. </p><p>Model collapse would promote hallucinations within LLMs, as well as loss of key training information, the potential for poor decision making, and degraded <a href="https://www.itpro.com/technology/artificial-intelligence/the-risks-of-open-source-ai-models">AI guardrails</a> that will erode business trust in affected AI models.</p><h2 id="the-three-stages-of-model-collapse">The three stages of model collapse  </h2><p>Models are undergoing partial synthetic training and for now, the threshold of real data tends to be higher than that of synthetic data. However, models can collapse when the percentage of synthetic data exceeds real data. There are three stages of model collapse:</p><p><strong>Early collapse</strong>: The model starts to skip over or 'forget' some rare, uncommon, and unusual information. Early collapse is less likely to be noticed because model runs still generate quality information.</p><p><strong>Late collapse</strong>: this marks the loss of structure in synthetic data. The model blends different patterns to generate outputs that may sound similar, but don't resemble real-world situations. A model undergoing late collapse produces vague and low-quality responses. </p><p><strong>Total collapse</strong>: Total collapse is a hypothetical event that occurs when a very high threshold of synthetic data enters the training environment. The model enters into a “self-consuming loop” and relies only on synthetic data.    </p><iframe allow="" height="200px" width="100%" id="" style="" class="position-center" data-lazy-priority="low" data-lazy-src="https://player.captivate.fm/episode/db98e7d4-39e3-4dfe-bdfa-e05ab9c11b6d/" src="https://player.captivate.fm/episode/db98e7d4-39e3-4dfe-bdfa-e05ab9c11b6d/"></iframe><h2 id="errors-causing-the-collapse">Errors causing the collapse</h2><p>Errors in training data lead to model collapse and the size of the sampling data influences error rates. If enterprises use an infinite number of samples, errors would disappear. That’s how strange mathematics is. However, infinite data does not exist on the internet: it’s a physical impossibility.</p><p>This is why synthetic data, used correctly, is actually good for training quality AI models. The key is preparing the right mix of organic and synthetic data at every stage of the training process.</p><p>Every time a model learns from sampled data, some approximation error occurs. When a large volume of synthetic data enters the training cycle, statistical errors appear more frequently. As a result, small statistical errors integrate over time, leading to model collapse.  </p><p>Statistical approximation errors are the primary reasons for model collapse. Models perform regression tasks, essentially learning to draw links between disparate data points. Errors are deviations between true values and approximations derived from models, which lead to models confidently outputting generalized or flat-out incorrect answers. A collapsing model would generate more deviations from true values. </p><p>All models assign weight to neurons in a neural network. The higher the weight, the higher the relevance of the dependent variable. But the data on which AI is trained is only ever an approximation of complex data – and due to functional approximation errors, models can draw incorrect links between training data, which leads to them generating false data outside the training data down the line. Functional approximation errors are the secondary cause of model collapse, which don’t grow with generations. </p><p><a href="https://www.itpro.com/security/hackers-are-deliberately-poisoning-ai-systems-to-make-them-malfunction-and-theres-no-way-to-defend-against-it"><u>Data poisoning</u></a> is another indirect reason for model collapse. In this cyber plan, hackers can intentionally insert misleading data into training loops. As a result, models may learn wrong patterns to generate false information. While this isn’t a major risk right now, it’s something that will have to be guarded against as future models are trained.</p><p>“Most organizations do not have enough high-quality data to train a model to the precision they need for good business outcomes, and few build foundation models from scratch,” says Christopher Royles, field CTO EMEA at Cloudera. </p><p>“The best way is to start with model selection,” he says, advocating for a firm business data strategy that deploys a blend of organic and synthetic data.</p><p>Even after selecting optimal training data, Royles says, it is impossible to prevent synthetic data from entering the loop.<em> </em>"Begin with real-world samples, labelled and enriched with <a href="https://www.itpro.com/development/web-development/368248/what-is-metadata"><u>metadata</u></a> that capture quality". Royles shares a step-wise breakdown to deal with model collapse. "Cluster the data to understand the statistical patterns within it, then generate synthetic records that mirror those patterns. LLM-as-a-judge can help here.” </p><p>Model collapse remains inevitable, but can be delayed if human-generated information maintains a higher percentage in training data. Thanks to years of big data. </p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/what-is-model-collapse-and-why-is-it-a-risk-for-enterprise-ai</link>
                                                                            <description>
                            <![CDATA[ Model collapse is a nightmare for AI companies and users, with AI models trained on AI data losing quality with each generation ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">gfErSxiQN9kmwtkk99yK95</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/Fi4V6sbqNLMKZi8fPe6VcL-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 10 Apr 2026 16:35:19 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:58 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Venus Kohli ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/MDHLkwNFLuDTZe75iZQLSU-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Venus is a freelance technology writer specializing in IT, quantum physics, electronics, and among other technical fields. She holds a degree in Electronics and Telecommunications Engineering from Mumbai University, India. &lt;/p&gt;&lt;p&gt;Alongside &lt;em&gt;ITPro&lt;/em&gt;, Venus has written for brands including &lt;em&gt;TechTarget, &lt;/em&gt;Kigen, Wevolver, and Narrato.&lt;/p&gt;&lt;p&gt;With years of experience in writing for global media brands and IT companies, she enjoys translating complex content into engaging stories. When she’s not writing about the latest IT trends, Venus can be found tracking  enterprise trends or the newest processor in town.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/Fi4V6sbqNLMKZi8fPe6VcL-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Visual art of green pixelated gradients, representing AI model collapse.]]></media:description>                                                            <media:text><![CDATA[Visual art of green pixelated gradients, representing AI model collapse.]]></media:text>
                                <media:title type="plain"><![CDATA[Visual art of green pixelated gradients, representing AI model collapse.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/Fi4V6sbqNLMKZi8fPe6VcL-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Ever since <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> became mainstream, AI generated content has become increasingly common. From text, to audio, images, and video, <a href="https://www.itpro.com/technology/artificial-intelligence/what-is-synthetic-data"><u>synthetic data</u></a> has become easier than ever to produce.</p><p>A study by <a href="https://ahrefs.com/blog/what-percentage-of-new-content-is-ai-generated/"><u>Ahrefs</u></a>, published in 2025, analyzed 900,000 English-language web pages. About 2.5% of pages were purely AI-generated, and 71.7% of pages were a mix of human-written and AI-generated.</p><p>It's clear the percentage of AI-generated content, along with mixed content, is only going to increase. In some aspects, this is great news for enterprises, which can generate large, sanitized datasets that closely mimic their specific enterprise data without exposing actual private information to AI firms.</p><p>“Synthetic data shouldn’t be about replacing real data entirely but instead enhancing and extending it to build great AI models and agents,” says Dael Williamson, EMEA CTO at Databricks.</p><p>Ironically, however, the explosion of synthetic data has problematic implications for training future generations of AI models. Some firms, like Gartner, are already <a href="https://www.itpro.com/security/data-protection/fears-over-ai-model-collapse-are-fueling-a-shift-to-zero-trust-data-governance-strategies"><u>sounding the alarm</u></a> over how the massive volumes of AI-generated data coming down the pipeline. The concern is training on AI data could create a sort of algorithmic feedback loop, in which AI models become increasingly distanced from real data and more prone to <a href="https://www.itpro.com/technology/artificial-intelligence/ai-hallucinations-what-are-they"><u>hallucinations</u></a>. </p><p>A blend of intelligent AI-generated and human-written information is important. What if AI-generated information exhausts the percentage of real content, reaching its critical limit?</p><h2 id="what-is-model-collapse">What is model collapse?</h2><p>In a <a href="https://www.nature.com/articles/s41586-024-07566-y"><u>joint study</u></a>, Ilia Shumailov from the University of Oxford and Zakhar Shumaylov coined the term “model collapse”. </p><p>Model collapse is defined as a long-term statistical degenerative process in which an AI model gradually misperceives reality, bets on improbable events, and starts to generate low-quality or wrong outputs. An AI model begins to collapse when its training data gets contaminated by its output or other AI-generated information.</p><p>For example, model A trains on purely human-generated information. Model A now generates a large volume of synthetic data. A new model B now uses this synthetic data as a part of its training. Then, model C trains on synthetic data from models A and B… and so on. That’s how synthetic data ruins the training data distribution.</p><p>Anyone who ever tried to copy a VHS tape more than once has experienced generation loss, in which each reproduction of data produces a lesser copy filled with more and more artifacts. Model collapse could happen in much the same way, with AI developers forced to run more and more costly training runs to eke out performance increases in subsequent generations of frontier AI models.</p><p>Model collapse isn’t limited to LLMs: <a href="https://www.itpro.com/network-internet/29791/what-is-an-artificial-neural-network"><u>neural network</u></a> architectures such as variational encoders and probabilistic models such as mixture models are also susceptible.</p><p>A real world example can be found in OPT-125M, an <a href="https://www.itpro.com/technology/artificial-intelligence/three-open-source-large-language-models-you-can-use-today"><u>open large language model</u></a> created by Meta AI researchers. To test model collapse, the researchers trained OPT-125M on its own outputs, then prompted it to answer questions on historical architecture. In response, the model produced answers rabbits instead. In the near future, the nth generation of AI models would have to fight just such a collapse.</p><p>Williamson uses a real-world example to explain model collapse: </p><p>"Imagine a locksmith who keeps duplicating keys from the latest copy, not the original. Early on, the key still turns most locks.</p><p>“But after many rounds, the key may look right at a glance, yet it jams, scrapes, or spins uselessly. Copying copies collapses fidelity, but copying while anchored to the master preserves it”,<em> </em>Williamson adds. </p><p>Model collapse would promote hallucinations within LLMs, as well as loss of key training information, the potential for poor decision making, and degraded <a href="https://www.itpro.com/technology/artificial-intelligence/the-risks-of-open-source-ai-models">AI guardrails</a> that will erode business trust in affected AI models.</p><h2 id="the-three-stages-of-model-collapse">The three stages of model collapse  </h2><p>Models are undergoing partial synthetic training and for now, the threshold of real data tends to be higher than that of synthetic data. However, models can collapse when the percentage of synthetic data exceeds real data. There are three stages of model collapse:</p><p><strong>Early collapse</strong>: The model starts to skip over or 'forget' some rare, uncommon, and unusual information. Early collapse is less likely to be noticed because model runs still generate quality information.</p><p><strong>Late collapse</strong>: this marks the loss of structure in synthetic data. The model blends different patterns to generate outputs that may sound similar, but don't resemble real-world situations. A model undergoing late collapse produces vague and low-quality responses. </p><p><strong>Total collapse</strong>: Total collapse is a hypothetical event that occurs when a very high threshold of synthetic data enters the training environment. The model enters into a “self-consuming loop” and relies only on synthetic data.    </p><iframe allow="" height="200px" width="100%" id="" style="" class="position-center" data-lazy-priority="low" data-lazy-src="https://player.captivate.fm/episode/db98e7d4-39e3-4dfe-bdfa-e05ab9c11b6d/" src="https://player.captivate.fm/episode/db98e7d4-39e3-4dfe-bdfa-e05ab9c11b6d/"></iframe><h2 id="errors-causing-the-collapse">Errors causing the collapse</h2><p>Errors in training data lead to model collapse and the size of the sampling data influences error rates. If enterprises use an infinite number of samples, errors would disappear. That’s how strange mathematics is. However, infinite data does not exist on the internet: it’s a physical impossibility.</p><p>This is why synthetic data, used correctly, is actually good for training quality AI models. The key is preparing the right mix of organic and synthetic data at every stage of the training process.</p><p>Every time a model learns from sampled data, some approximation error occurs. When a large volume of synthetic data enters the training cycle, statistical errors appear more frequently. As a result, small statistical errors integrate over time, leading to model collapse.  </p><p>Statistical approximation errors are the primary reasons for model collapse. Models perform regression tasks, essentially learning to draw links between disparate data points. Errors are deviations between true values and approximations derived from models, which lead to models confidently outputting generalized or flat-out incorrect answers. A collapsing model would generate more deviations from true values. </p><p>All models assign weight to neurons in a neural network. The higher the weight, the higher the relevance of the dependent variable. But the data on which AI is trained is only ever an approximation of complex data – and due to functional approximation errors, models can draw incorrect links between training data, which leads to them generating false data outside the training data down the line. Functional approximation errors are the secondary cause of model collapse, which don’t grow with generations. </p><p><a href="https://www.itpro.com/security/hackers-are-deliberately-poisoning-ai-systems-to-make-them-malfunction-and-theres-no-way-to-defend-against-it"><u>Data poisoning</u></a> is another indirect reason for model collapse. In this cyber plan, hackers can intentionally insert misleading data into training loops. As a result, models may learn wrong patterns to generate false information. While this isn’t a major risk right now, it’s something that will have to be guarded against as future models are trained.</p><p>“Most organizations do not have enough high-quality data to train a model to the precision they need for good business outcomes, and few build foundation models from scratch,” says Christopher Royles, field CTO EMEA at Cloudera. </p><p>“The best way is to start with model selection,” he says, advocating for a firm business data strategy that deploys a blend of organic and synthetic data.</p><p>Even after selecting optimal training data, Royles says, it is impossible to prevent synthetic data from entering the loop.<em> </em>"Begin with real-world samples, labelled and enriched with <a href="https://www.itpro.com/development/web-development/368248/what-is-metadata"><u>metadata</u></a> that capture quality". Royles shares a step-wise breakdown to deal with model collapse. "Cluster the data to understand the statistical patterns within it, then generate synthetic records that mirror those patterns. LLM-as-a-judge can help here.” </p><p>Model collapse remains inevitable, but can be delayed if human-generated information maintains a higher percentage in training data. Thanks to years of big data. </p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Microsoft says Copilot isn't just for 'entertainment purposes only' ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Microsoft has confirmed that terms describing Copilot as for  “entertainment purposes only" need to be updated.  </p><p>The terms of use for Copilot have raised eyebrows, sparking bemusement online and coming amid wider concerns that the use of AI in businesses is leading to <a href="https://www.itpro.com/technology/artificial-intelligence/workers-are-wasting-half-a-day-each-week-fixing-ai-workslop"><u>"workslop" that employees</u></a> spend half their day fixing. </p><p>The terms state: "Copilot is for entertainment purposes only. It can make mistakes, and it may not work as intended. Don’t rely on Copilot for important advice. Use Copilot at your own risk."</p><p>That <a href="https://www.microsoft.com/en-us/microsoft-copilot/for-individuals/termsofuse" target="_blank"><u>document</u></a> was last updated back in October, when they were "rewritten and reorganized." </p><p>Previous versions of the terms stretching back to 2023 didn't include that exact same line, but did note: "The Online Services are for entertainment purposes; the Online Services are not error-free, may not work as expected and may generate incorrect information."</p><h2 id="microsoft-confirms-t-cs-changes">Microsoft confirms T&amp;Cs changes</h2><p>Though the language isn't entirely new, it was spotted by users and posted to social media at a time when Microsoft is facing some backlash to how heavily it's embedding Copilot into Windows. </p><p>Last month <a href="https://www.itpro.com/software/windows/microsoft-is-rolling-back-unnecessary-copilot-features-on-windows"><u>Microsoft announced plans to dial back</u></a> some AI features in Windows amid a growing backlash against Copilot, and in <a href="https://www.itpro.com/software/not-keen-on-microsoft-copilot-dont-worry-your-admins-can-now-uninstall-it-but-only-if-youve-not-used-it-within-28-days"><u>January revealed plans</u></a> to allow admins to remove the Copilot app completely if it wasn't being used. </p><p>The tech giant has also recorded its worst quarter since the 2008 financial crisis, with one analyst <a href="https://www.cnbc.com/2026/03/31/microsofts-stock-closes-worst-quarter-since-2008-financial-crisis.html" target="_blank">suggesting </a>the firm is "in a pickle" over issues with Copilot, particularly adoption rates. </p><p>Microsoft said it would update the disclaimer as the terms are now out of date. </p><p>"The ‘entertainment purposes’ phrasing is legacy language from when Copilot originally launched as a search companion service in Bing," a Microsoft spokesperson told <a href="https://uk.pcmag.com/ai/164230/copilot-terms-claim-microsofts-ai-is-for-entertainment-purposes-only" target="_blank"><u><em>PCMag</em></u></a>. </p><p>"As the product has evolved, that language is no longer reflective of how Copilot is used today and will be altered with our next update." </p><p><em>ITPro </em>approached Microsoft for comment, but did not receive a response by time of publication. </p><h2 id="ai-warnings">AI warnings</h2><p>While <a href="https://uk.pcmag.com/ai/164230/copilot-terms-claim-microsofts-ai-is-for-entertainment-purposes-only" target="_blank"><u>reports</u></a> noted social media <a href="https://www.reddit.com/r/pcmasterrace/comments/1s8u3t7/direct_quote_from_microsofts_terms_of_use_for/odjgv10/" target="_blank"><u>users</u></a> laughing that Microsoft's language echoes the legalese used by psychics and ghost hunters, such disclaimers are also common among AI developers — though they warn about potential risks rather than declaring their products "entertainment". </p><p>The <a href="https://x.ai/legal/terms-of-service" target="_blank"><u>terms for xAI</u></a> state that its service "may contain errors, defects, bugs or inaccuracies that could fail or cause corruption or loss of data and information”. </p><p>These particular terms request users agree not to "rely on output as the sole source of truth or factual information, or as professional advice." </p><p>OpenAI features the same language, also warning the service may not be "error free" while Anthropic warned that Claude outputs “may not always be accurate" and "actions may not be error free".</p><p>Microsoft also adds in its terms: "always use your judgment and check the information you get from Copilot before you make decisions or act."</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/microsoft-plans-tweaks-to-copilot-terms-claiming-chatbot-is-for-entertainment-purposes-only</link>
                                                                            <description>
                            <![CDATA[ Sharp-eyed users spotted Microsoft describing its Copilot AI as "for entertainment purposes only" ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">4hZqiAs6XJEw9v6G2LaAv3</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/6FdBRNc4i8SUjVx5Qf5qkR-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 07 Apr 2026 09:35:36 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:42 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/6FdBRNc4i8SUjVx5Qf5qkR-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Microsoft Copilot logo and branding pictured on a smartphone screen, with smartphone placed on top of a laptop keyboard.]]></media:description>                                                            <media:text><![CDATA[Microsoft Copilot logo and branding pictured on a smartphone screen, with smartphone placed on top of a laptop keyboard.]]></media:text>
                                <media:title type="plain"><![CDATA[Microsoft Copilot logo and branding pictured on a smartphone screen, with smartphone placed on top of a laptop keyboard.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/6FdBRNc4i8SUjVx5Qf5qkR-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Microsoft has confirmed that terms describing Copilot as for  “entertainment purposes only" need to be updated.  </p><p>The terms of use for Copilot have raised eyebrows, sparking bemusement online and coming amid wider concerns that the use of AI in businesses is leading to <a href="https://www.itpro.com/technology/artificial-intelligence/workers-are-wasting-half-a-day-each-week-fixing-ai-workslop"><u>"workslop" that employees</u></a> spend half their day fixing. </p><p>The terms state: "Copilot is for entertainment purposes only. It can make mistakes, and it may not work as intended. Don’t rely on Copilot for important advice. Use Copilot at your own risk."</p><p>That <a href="https://www.microsoft.com/en-us/microsoft-copilot/for-individuals/termsofuse" target="_blank"><u>document</u></a> was last updated back in October, when they were "rewritten and reorganized." </p><p>Previous versions of the terms stretching back to 2023 didn't include that exact same line, but did note: "The Online Services are for entertainment purposes; the Online Services are not error-free, may not work as expected and may generate incorrect information."</p><h2 id="microsoft-confirms-t-cs-changes">Microsoft confirms T&amp;Cs changes</h2><p>Though the language isn't entirely new, it was spotted by users and posted to social media at a time when Microsoft is facing some backlash to how heavily it's embedding Copilot into Windows. </p><p>Last month <a href="https://www.itpro.com/software/windows/microsoft-is-rolling-back-unnecessary-copilot-features-on-windows"><u>Microsoft announced plans to dial back</u></a> some AI features in Windows amid a growing backlash against Copilot, and in <a href="https://www.itpro.com/software/not-keen-on-microsoft-copilot-dont-worry-your-admins-can-now-uninstall-it-but-only-if-youve-not-used-it-within-28-days"><u>January revealed plans</u></a> to allow admins to remove the Copilot app completely if it wasn't being used. </p><p>The tech giant has also recorded its worst quarter since the 2008 financial crisis, with one analyst <a href="https://www.cnbc.com/2026/03/31/microsofts-stock-closes-worst-quarter-since-2008-financial-crisis.html" target="_blank">suggesting </a>the firm is "in a pickle" over issues with Copilot, particularly adoption rates. </p><p>Microsoft said it would update the disclaimer as the terms are now out of date. </p><p>"The ‘entertainment purposes’ phrasing is legacy language from when Copilot originally launched as a search companion service in Bing," a Microsoft spokesperson told <a href="https://uk.pcmag.com/ai/164230/copilot-terms-claim-microsofts-ai-is-for-entertainment-purposes-only" target="_blank"><u><em>PCMag</em></u></a>. </p><p>"As the product has evolved, that language is no longer reflective of how Copilot is used today and will be altered with our next update." </p><p><em>ITPro </em>approached Microsoft for comment, but did not receive a response by time of publication. </p><h2 id="ai-warnings">AI warnings</h2><p>While <a href="https://uk.pcmag.com/ai/164230/copilot-terms-claim-microsofts-ai-is-for-entertainment-purposes-only" target="_blank"><u>reports</u></a> noted social media <a href="https://www.reddit.com/r/pcmasterrace/comments/1s8u3t7/direct_quote_from_microsofts_terms_of_use_for/odjgv10/" target="_blank"><u>users</u></a> laughing that Microsoft's language echoes the legalese used by psychics and ghost hunters, such disclaimers are also common among AI developers — though they warn about potential risks rather than declaring their products "entertainment". </p><p>The <a href="https://x.ai/legal/terms-of-service" target="_blank"><u>terms for xAI</u></a> state that its service "may contain errors, defects, bugs or inaccuracies that could fail or cause corruption or loss of data and information”. </p><p>These particular terms request users agree not to "rely on output as the sole source of truth or factual information, or as professional advice." </p><p>OpenAI features the same language, also warning the service may not be "error free" while Anthropic warned that Claude outputs “may not always be accurate" and "actions may not be error free".</p><p>Microsoft also adds in its terms: "always use your judgment and check the information you get from Copilot before you make decisions or act."</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Satya Nadella needs to remember the Streisand effect for 'AI slop' ]]></title>
                                                                                                <dc:content><![CDATA[ <p>It’s not often that I feel bad for the chief executive of a multinational, multi-trillion dollar company. Well, when Satya Nadella, CEO at Microsoft, wrote a LinkedIn post in late 2025 all but begging the public to “...<a href="https://www.itpro.com/technology/artificial-intelligence/satya-nadella-microsoft-ai-slop-2026"><u>get beyond the arguments of slop</u></a> vs sophistication and develop a new equilibrium in terms of our ‘theory of the mind’”, I’m happy to say he kept my streak alive. </p><p>He also doubled down, like most tech leaders are want to do these days, at the <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-ceo-satya-nadella-ai-bubble-claims"><u>World Economic Forum in Davos</u></a> where he <a href="https://www.politico.com/news/2025/12/01/microsofts-nadella-says-ai-must-earn-social-permission-to-consume-so-much-energy-00671920"><u>said</u></a> AI needed broader adoption because of a possible loss of “social permission” to use natural resources if it fails and gave a fairly dire definition of an AI bubble. </p><p>“I mean I think a telltale sign of if it's a bubble would be if all we're talking about are the tech firms,” Nadella added.</p><p>I take some solace in the fact that the same company quasi-demanding that the <a href="https://www.itpro.com/software/microsoft/microsofts-copilot-keyboard-button-might-be-the-best-way-to-drive-widespread-adoption-of-its-beloved-ai-features"><u>Copilot key</u></a> on your keyboard be mandatory is entering the spring of their discontent. Two massive branding problems and we’re not even halfway through the year.</p><p>You see, Nadella failed to heed the lessons of what is commonly known as the Streisand Effect. This is the phenomenon in the court of public opinion, so often broadcast on social media, where trying to actively get people to ignore something you’ve done will actually cause the reverse to happen. As <em>ITPro’s</em> managing editor, Jane McCallion, put it on <a href="https://bsky.app/profile/janemccallion.bsky.social/post/3mbok6z63rc2t" target="_blank"><u>Bluesky</u></a>: “I can't quite put my finger on why, but Nadella requesting we stop saying AI slop reminds me of Jeb Bush asking the audience to ‘please clap’.” </p><p>While Nadella’s approach wasn’t quite as meme-tastic as Bush’s faux pas – at least Bush didn’t try to dress up his wording as C-Suite style optimism – the underlying intent is rooted in the same thing: an undying belief that those with the biggest budgets can make magic in public. His blog post is the first crack in this AI-powered armor.</p><p>What Nadella and his colleagues have failed to understand is rule number one of public relations: don’t fan the flames. One can just imagine the seismic amount of headslapping that went on in Microsoft’s PR offices.</p><p>The good news for Nadella is that he’s far from the first tech CEO to stick his foot in his mouth in this fashion. Hell, there were leaders at <a href="https://www.itpro.com/technology/live/ces-2026-all-the-live-updates-as-they-happen-from-day-one"><u>CES in January</u></a> that made similarly dubious announcements. It’s hard to defend, for example, companies appearing at the convention while simultaneously <a href="https://www.channelnews.com.au/samsung-exits-show-floor-at-ces-as-they-go-after-new-b2b-markets/" target="_blank"><u>pulling out of the consumer vertical</u></a> for their own products and nothing quite says AI slop as valiantly as <a href="https://www.youtube.com/shorts/FFyamyoq4cI" target="_blank"><u>Razer’s AI anime companion</u></a>. </p><p>So, allow me to take a few lessons from that theatre degree, focused on disability, identity, and solo performance, and impart them to tech leadership who only see the arts as a good charitable write off. </p><p>First, take a breath. It should surprise no one that upper level executives at massive tech companies could be out of touch with the general public, but I think the root of the issue is they don’t take more than a few seconds in those very public moments to remember who their audience is. </p><p>On stage at Davos, BlackRock CEO Larry Fink was a friendly interviewer for Nadella and no doubt the audience was similarly inclined. The key difference is that messages at the World Economic Forum have a different audience than shareholders meetings. These comments will break containment. The everyday person Nadella kept talking about is not, as he posits, the knowledge worker or healthcare administrator crying out for efficiency; they are people with <a href="https://institute.global/insights/tech-and-digitalisation/what-the-uk-thinks-about-ai-building-public-trust-to-accelerate-adoption"><u>growing distrust</u></a> of the content these AI models are spitting out. </p><p>Second, go back to basics. Regardless of whether we’re talking about art, sports, or tech, when you start to lose the thread, it’s time to come back to your foundation. </p><p>Not some self indulgent mantra of CEO first principles but a grounded process, aimed at key people in your user base. If a basketball player is in a shooting slump, they’ll often use a deconstructed form to get back in rhythm; a singer has a really specific warm up and cool down; and even someone on Nadella’s scale could use going back to these simple reps. Ditch the giant corporate case studies and look at what “social permission” actually means. </p><p>Lastly, come up for air. When you’re in a theatre show it can often be really easy to stay within your bubble. There’s this sort of post-performance malaise you can fall into once a show is over, as you come out of this creative cocoon and enter back into the real world. </p><p>Now, I’m not insinuating that the bear pit of multinational corporations is anything like an indie theatre rehearsal environment. What I can tell you is this: both situations require the fresh air of a reset. Nadella and his compatriots could use a dose of time away from these self indulgent spaces like the World Economic Forum or a self-penned blog post. </p><p>Otherwise, we’ll keep seeing these headlines. But hey, at least the Reddit content is good for a laugh.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/satya-nadella-needs-to-remember-the-streisand-effect-for-ai-slop</link>
                                                                            <description>
                            <![CDATA[ Attempts to discourage criticism may backfire for Microsoft’s CEO ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">ZcsWH7wenSXUfEzXJUvYBo</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/Ssf2zEL2sCjDwUNYdRVRmn-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 24 Mar 2026 13:32:46 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:42 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ john@jloeppky.com (John Loeppky) ]]></author>                    <dc:creator><![CDATA[ John Loeppky ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GJCxqX7ryKSC5XjEDLnEtU-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;John Loeppky is a British-Canadian disabled freelance writer currently based on Treaty 6 territory in Saskatoon, Saskatchewan, Canada where he lives with his wife and three dogs.&lt;/p&gt;&lt;p&gt;In addition to his work for ITPro, he regularly works with outlets such as &lt;em&gt;CBC&lt;/em&gt;, &lt;em&gt;Healthline&lt;/em&gt;, &lt;em&gt;VeryWell&lt;/em&gt;, &lt;em&gt;Defector&lt;/em&gt;, and a host of others. John began his journalism career at the University of Regina at their student newspaper, The Carillon. He started as a sports writer, then sports editor, before serving as EIC until 2020. &lt;/p&gt;&lt;p&gt;John holds a BA and an MFA from the University of Regina. His graduate work focused on disability, identity, and solo performance in theatre. Prior to shifting to being a full-time freelancer, John worked for Listen to Dis&#039; Community Arts Organization, Saskatchewan&#039;s only disability-led disability arts organization. &lt;/p&gt;&lt;p&gt;His focus as a generalist is to cover things with societal and cultural impact, particularly when it comes to inclusion in its various forms. As a result, he&#039;s written for FiveThirtyEight about Apple Watches, covered the Paris Paralympics for CBC, and written for Defector about a scandal in American wheelchair basketball. He also serves as a member of the National Center on Disability and Journalism&#039;s advisory board. John&#039;s goal in life is to have an entertaining obituary to read. &lt;/p&gt;&lt;p&gt;You can find more of his work at Jloeppky.com/portfolio and you can contact him via email at John@Jloeppky.com.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/Ssf2zEL2sCjDwUNYdRVRmn-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A telephoto shot of Satya Nadella, CEO at Microsoft, speaking onstage at the World Economic Forum 2026 in Davos.]]></media:description>                                                            <media:text><![CDATA[A telephoto shot of Satya Nadella, CEO at Microsoft, speaking onstage at the World Economic Forum 2026 in Davos.]]></media:text>
                                <media:title type="plain"><![CDATA[A telephoto shot of Satya Nadella, CEO at Microsoft, speaking onstage at the World Economic Forum 2026 in Davos.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/Ssf2zEL2sCjDwUNYdRVRmn-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>It’s not often that I feel bad for the chief executive of a multinational, multi-trillion dollar company. Well, when Satya Nadella, CEO at Microsoft, wrote a LinkedIn post in late 2025 all but begging the public to “...<a href="https://www.itpro.com/technology/artificial-intelligence/satya-nadella-microsoft-ai-slop-2026"><u>get beyond the arguments of slop</u></a> vs sophistication and develop a new equilibrium in terms of our ‘theory of the mind’”, I’m happy to say he kept my streak alive. </p><p>He also doubled down, like most tech leaders are want to do these days, at the <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-ceo-satya-nadella-ai-bubble-claims"><u>World Economic Forum in Davos</u></a> where he <a href="https://www.politico.com/news/2025/12/01/microsofts-nadella-says-ai-must-earn-social-permission-to-consume-so-much-energy-00671920"><u>said</u></a> AI needed broader adoption because of a possible loss of “social permission” to use natural resources if it fails and gave a fairly dire definition of an AI bubble. </p><p>“I mean I think a telltale sign of if it's a bubble would be if all we're talking about are the tech firms,” Nadella added.</p><p>I take some solace in the fact that the same company quasi-demanding that the <a href="https://www.itpro.com/software/microsoft/microsofts-copilot-keyboard-button-might-be-the-best-way-to-drive-widespread-adoption-of-its-beloved-ai-features"><u>Copilot key</u></a> on your keyboard be mandatory is entering the spring of their discontent. Two massive branding problems and we’re not even halfway through the year.</p><p>You see, Nadella failed to heed the lessons of what is commonly known as the Streisand Effect. This is the phenomenon in the court of public opinion, so often broadcast on social media, where trying to actively get people to ignore something you’ve done will actually cause the reverse to happen. As <em>ITPro’s</em> managing editor, Jane McCallion, put it on <a href="https://bsky.app/profile/janemccallion.bsky.social/post/3mbok6z63rc2t" target="_blank"><u>Bluesky</u></a>: “I can't quite put my finger on why, but Nadella requesting we stop saying AI slop reminds me of Jeb Bush asking the audience to ‘please clap’.” </p><p>While Nadella’s approach wasn’t quite as meme-tastic as Bush’s faux pas – at least Bush didn’t try to dress up his wording as C-Suite style optimism – the underlying intent is rooted in the same thing: an undying belief that those with the biggest budgets can make magic in public. His blog post is the first crack in this AI-powered armor.</p><p>What Nadella and his colleagues have failed to understand is rule number one of public relations: don’t fan the flames. One can just imagine the seismic amount of headslapping that went on in Microsoft’s PR offices.</p><p>The good news for Nadella is that he’s far from the first tech CEO to stick his foot in his mouth in this fashion. Hell, there were leaders at <a href="https://www.itpro.com/technology/live/ces-2026-all-the-live-updates-as-they-happen-from-day-one"><u>CES in January</u></a> that made similarly dubious announcements. It’s hard to defend, for example, companies appearing at the convention while simultaneously <a href="https://www.channelnews.com.au/samsung-exits-show-floor-at-ces-as-they-go-after-new-b2b-markets/" target="_blank"><u>pulling out of the consumer vertical</u></a> for their own products and nothing quite says AI slop as valiantly as <a href="https://www.youtube.com/shorts/FFyamyoq4cI" target="_blank"><u>Razer’s AI anime companion</u></a>. </p><p>So, allow me to take a few lessons from that theatre degree, focused on disability, identity, and solo performance, and impart them to tech leadership who only see the arts as a good charitable write off. </p><p>First, take a breath. It should surprise no one that upper level executives at massive tech companies could be out of touch with the general public, but I think the root of the issue is they don’t take more than a few seconds in those very public moments to remember who their audience is. </p><p>On stage at Davos, BlackRock CEO Larry Fink was a friendly interviewer for Nadella and no doubt the audience was similarly inclined. The key difference is that messages at the World Economic Forum have a different audience than shareholders meetings. These comments will break containment. The everyday person Nadella kept talking about is not, as he posits, the knowledge worker or healthcare administrator crying out for efficiency; they are people with <a href="https://institute.global/insights/tech-and-digitalisation/what-the-uk-thinks-about-ai-building-public-trust-to-accelerate-adoption"><u>growing distrust</u></a> of the content these AI models are spitting out. </p><p>Second, go back to basics. Regardless of whether we’re talking about art, sports, or tech, when you start to lose the thread, it’s time to come back to your foundation. </p><p>Not some self indulgent mantra of CEO first principles but a grounded process, aimed at key people in your user base. If a basketball player is in a shooting slump, they’ll often use a deconstructed form to get back in rhythm; a singer has a really specific warm up and cool down; and even someone on Nadella’s scale could use going back to these simple reps. Ditch the giant corporate case studies and look at what “social permission” actually means. </p><p>Lastly, come up for air. When you’re in a theatre show it can often be really easy to stay within your bubble. There’s this sort of post-performance malaise you can fall into once a show is over, as you come out of this creative cocoon and enter back into the real world. </p><p>Now, I’m not insinuating that the bear pit of multinational corporations is anything like an indie theatre rehearsal environment. What I can tell you is this: both situations require the fresh air of a reset. Nadella and his compatriots could use a dose of time away from these self indulgent spaces like the World Economic Forum or a self-penned blog post. </p><p>Otherwise, we’ll keep seeing these headlines. But hey, at least the Reddit content is good for a laugh.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Are small language models finally having their moment? ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Large language models (LLMs) with billions, or even trillions, of parameters come with a baggage of massive training budgets, <a href="https://www.itpro.com/infrastructure/ai-industry-copper-supply-shortages-infrastructure-challenges"><u>material shortages</u></a>, <a href="https://www.itpro.com/technology/artificial-intelligence/the-costs-of-building-generative-ai-platforms-are-racking-up"><u>processing costs</u></a>, and <a href="https://www.itpro.com/software/development/ai-generated-code-is-fast-becoming-the-biggest-enterprise-security-risk-as-teams-struggle-with-the-illusion-of-correctness"><u>security risks</u></a>. Despite these limitations, LLMs have been widely adopted across enterprises – a <a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai" target="_blank"><u>recent report</u></a> from McKinsey and company found 88% of organizations now use AI in at least one business function. </p><p>In most enterprises, data security, compliance, and intellectual property concerns surface in boardroom discussions. Sharing proprietary data on the cloud with externally hosted LLMs restricted enterprises from using LLMs. </p><p>That’s where <a href="https://www.itpro.com/technology/artificial-intelligence/small-language-models-are-growing-in-popularity-but-they-have-a-hidden-fallacy-that-enterprises-must-come-to-terms-with"><u>small language models (SLMs)</u></a> fit into enterprise in-house operations. SLMs are compact AI models with fewer parameters, typically ranging from millions to a few billion. SLMs don’t rely on the internet to run and store data. There is no risk of a security breach due to the absence of cloud round-trips. The critical data generated on-premises, without the need for internet connectivity, ensures regulatory compliance and security.</p><p>There are several enterprise use cases of SLMs. According to a report by <a href="https://www.marketsandmarkets.com/Market-Reports/small-language-model-market-4008452.html" target="_blank"><u>Markets and Markets</u></a>, about 44% of SLM enterprise adoption is driven by the cloud. Enterprises follow a <a href="https://www.itpro.com/hybrid-cloud/29668/what-is-hybrid-cloud">hybrid cloud</a> strategy to lower hardware infrastructure costs.  </p><p>Gartner <a href="https://www.gartner.com/en/newsroom/press-releases/2025-04-09-gartner-predicts-by-2027-organizations-will-use-small-task-specific-ai-models-three-times-more-than-general-purpose-large-language-models" target="_blank">projects</a> that by 2027, organizations will use task-specific SLMs three times more than LLMs.</p><p>SLMs don’t replace LLMs but complement them. The same enterprise may choose SLMs for some secure tasks and LLMs for broader operations. Matt Beucler, CEO and founder of Plura AI, compares LLMs and SLMs.<em> </em>“SLMs handle speed, control, and security", says Beucler. “LLMs do abstraction, synthesis, and creativity. That’s an important conceptual dividing line in modern engineering, and companies that recognize it tend to build scalable, reliable systems.”</p><p>LLMs, Beucler tells <em>ITPro, </em>are also not brilliant at rules-based work on their own. This, and the lure of deploying more industry or case-specific AI models, is part of the appeal of SLMs. In most cases, SLMs are fine-tuned on domain knowledge to perform enterprise-specific tasks. For example, an SLM trained only on medical guidelines and patient data, rather than mathematical equations, would deliver context-aware outputs and exhibit lower chances of hallucinations. </p><p>SLM adoption is also largely driven by training and processing costs. The cost of training SLMs on a smaller dataset is significantly lower than that of LLMs. In 2024, <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-openais-gpt-4o-updates-including-pricing-changes-and-new-features"><u><em>ITPro</em></u><u> reported</u></a> that OpenAI’s GPT-4o mini was 60% cheaper than LLM GPT-3.5 Turbo. In terms of processing costs, LLMs may account for several tens of thousands of dollars across individual tasks, while SLMs reach just a fraction of that.</p><p>The training window of SLMs is as small as a week. LLMs, on the other hand, need months of training to achieve comparable results. SLMs are also less energy-intensive, designed to run locally on low-power devices. As a result, SLMs are an enterprise-favoured choice for <a href="https://www.itpro.com/technology/artificial-intelligence/why-edge-ai-has-significant-business-potential">edge AI</a> and embedded applications. In simple terms, SLMs offer resource efficiency and enhanced cybersecurity at a lower price compared to LLMs.</p><p>However, many SLMs are also derived from a process known as <a href="https://www.itpro.com/technology/artificial-intelligence/google-says-hacker-groups-are-using-gemini-to-augment-attacks-and-companies-are-even-stealing-its-models"><u>‘knowledge distillation’</u></a>, in which a ‘student’ model is trained by a larger model such as the trillion-parameter frontier models. This is important because it shows SLMs aren't a replacement for LLMs altogether.</p><h2 id="use-cases-for-slms">Use cases for SLMs</h2><p>SLM training involves pruning, a model compression technique to remove unnecessary weights from the data. As a result, SLMs fail to answer general and out-of-context queries, making them more useful for specialized applications.</p><p>The developers behind some larger models have released SLMs designed for very specific workloads. For example, in February Alibaba <a href="https://qwen.ai/blog?id=qwen3-coder-next" target="_blank"><u>released</u></a> Qwen-3-Coder-Next, an open-weight model designed for coding agents to run on a single modern Nvidia <a href="https://www.itpro.com/hardware/30399/what-is-a-gpu">GPU</a>.</p><p>Indeed, SLMs are often chosen for code and text generation capabilities. <em>ITPro</em> spoke to Peter Schneider, the senior product manager at Qt Group, a Finnish software company specializing in cross-platform development frameworks, to hear more about practical SLM deployment.</p><p>Schneider shared how they have worked with LLMs and their “student model” SLMs. “We’ve tested code completion on our Qt Modelling Language using several small language models like<strong> </strong><a href="https://www.itpro.com/technology/artificial-intelligence/openai-open-weight-ai-models">GPT-oss-20b</a> or Llama 3.3 70B against larger models like <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform">Gemini 3 Pro</a> Preview, <a href="https://www.itpro.com/technology/artificial-intelligence/openai-thought-it-hit-a-home-run-with-gpt-5-users-werent-so-keen">GPT-5</a>, and <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner">Claude 4.5</a>.”</p><p>“Without tuning, the choice is generally this: do you want a roughly 85% chance of a good code suggestion, or are you ok with about 50%? If you want the former, you go with a frontier model like Gemini or Claude. If you want the latter, then it's a royalty-free small model”.</p><p>Schneider stresses that this doesn’t mean that SLMs should be considered inherently worse options than larger models.</p><p>“That is not to say SLMs can't be good, or even great, for very specific use cases like code completion after you fine-tune the model. We have had success fine-tuning small models like Code Llama to be as good at code completion as bigger models.”</p><p>Defense and aerospace industries deploy SLMs on-prem to handle and analyze classified data. The healthcare industry deploys SLMs for data compliance and protection of patient history. Thanks to the ability of SLMs to function without the internet in case of an emergency. </p><p>“For customers who will never connect to a public cloud in highly regulated industries like medical and defense, the choice is pretty clear”, Schneider highlighted the important enterprise use case of SLMs.</p><p>Searching through data held on premises, or generating documents grounded in information held in a private cloud, is in fact an ideal use case for SLMs. Provided they’re backed with the right hardware, such as a few on premises server racks or on suitably powerful staff devices, they can complete tasks securely and with low latency. </p><p><strong>“</strong>When classifying thousands of support tickets, extracting invoice data, or flagging compliance issues, an SLM trained on your own data will often outperform a general-purpose giant at a fraction of the cost, running in milliseconds and keeping data strictly on-premises,” says Barry James, the director of data and AI at Nucleo.</p><p>Other common SLM use cases include:</p><ul><li><strong>Telecom:</strong> Contact centers are the most common examples where SLMs handle customer data, such as interactive voice response menu (IVR), conversations, ticket routing, FAQs, and product-related queries. With the help of SLMs, enterprises can rely on low-latency volume interactions and use customer data to forecast and predict trends.</li><li><strong>Embedded systems</strong>: SLMs can run without the internet on local edge devices. SLMs are suitable for industrial IoT applications that allow devices to operate in constrained settings with utmost safety features.</li><li><strong>Factories</strong>: SLMs require fewer GPUs to operate. In industries like semiconductor fabrication, SLMs can help build an enterprise-centric knowledge base of common defects and manufacturing errors.</li><li><strong>Banking</strong>: Financial data is one of the most critical types of data in enterprises and banks alike. Regulatory authorities and standards govern bank data globally. The use of AI in such applications is less discussed. With on-premises AI deployment, European banks might be able to comply with the <a href="https://www.itpro.com/it-legislation/27814/what-is-gdpr-everything-you-need-to-know">General Data Protection Regulation (GDPR)</a>.</li></ul><h2 id="notable-slms">Notable SLMs</h2><p>In 2026, some of the most notable SLMs include Microsoft’s Phi-4 and Phi4--mini, Google’s Gemma 3 and Gemma 3n, OpenAI’s gpt-oss-120b and gpt-oss-20b, Alibaba’s Qwen 2.5, and Meta’s Llama 4.</p><p>These models can be run on hardware much smaller than you would find in a data center. Microsoft says Phi-4-mini can be run on a single PC with a minimum of 16GB RAM. Gemma 3n is designed to run on smartphones, while depending on the quantized model size (ranging from 270 million to 27 billion) Gemma 3 can run on a consumer-grade laptop or may require a top-of-the-line GPU.</p><p>As the demand for SLMs increases, AI labs may focus more on quantization and squeezing the most performance possible out of small models.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/are-small-language-models-finally-having-their-moment</link>
                                                                            <description>
                            <![CDATA[ The smallest AI models can be run on a mobile phone or laptop – what's the business potential? ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">27Y5nXr3xgmxv6hoD4wzEo</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/XUFTVxXHrTRAVBdujTYj6P-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Mar 2026 17:50:10 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:49 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Venus Kohli ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/MDHLkwNFLuDTZe75iZQLSU-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Venus is a freelance technology writer specializing in IT, quantum physics, electronics, and among other technical fields. She holds a degree in Electronics and Telecommunications Engineering from Mumbai University, India. &lt;/p&gt;&lt;p&gt;Alongside &lt;em&gt;ITPro&lt;/em&gt;, Venus has written for brands including &lt;em&gt;TechTarget, &lt;/em&gt;Kigen, Wevolver, and Narrato.&lt;/p&gt;&lt;p&gt;With years of experience in writing for global media brands and IT companies, she enjoys translating complex content into engaging stories. When she’s not writing about the latest IT trends, Venus can be found tracking  enterprise trends or the newest processor in town.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/XUFTVxXHrTRAVBdujTYj6P-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Artificial intelligence (AI) concept image showing a digitized cube with &#039;AI&#039; written on it resting on top of circuit boards.]]></media:description>                                                            <media:text><![CDATA[Artificial intelligence (AI) concept image showing a digitized cube with &#039;AI&#039; written on it resting on top of circuit boards.]]></media:text>
                                <media:title type="plain"><![CDATA[Artificial intelligence (AI) concept image showing a digitized cube with &#039;AI&#039; written on it resting on top of circuit boards.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/XUFTVxXHrTRAVBdujTYj6P-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Large language models (LLMs) with billions, or even trillions, of parameters come with a baggage of massive training budgets, <a href="https://www.itpro.com/infrastructure/ai-industry-copper-supply-shortages-infrastructure-challenges"><u>material shortages</u></a>, <a href="https://www.itpro.com/technology/artificial-intelligence/the-costs-of-building-generative-ai-platforms-are-racking-up"><u>processing costs</u></a>, and <a href="https://www.itpro.com/software/development/ai-generated-code-is-fast-becoming-the-biggest-enterprise-security-risk-as-teams-struggle-with-the-illusion-of-correctness"><u>security risks</u></a>. Despite these limitations, LLMs have been widely adopted across enterprises – a <a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai" target="_blank"><u>recent report</u></a> from McKinsey and company found 88% of organizations now use AI in at least one business function. </p><p>In most enterprises, data security, compliance, and intellectual property concerns surface in boardroom discussions. Sharing proprietary data on the cloud with externally hosted LLMs restricted enterprises from using LLMs. </p><p>That’s where <a href="https://www.itpro.com/technology/artificial-intelligence/small-language-models-are-growing-in-popularity-but-they-have-a-hidden-fallacy-that-enterprises-must-come-to-terms-with"><u>small language models (SLMs)</u></a> fit into enterprise in-house operations. SLMs are compact AI models with fewer parameters, typically ranging from millions to a few billion. SLMs don’t rely on the internet to run and store data. There is no risk of a security breach due to the absence of cloud round-trips. The critical data generated on-premises, without the need for internet connectivity, ensures regulatory compliance and security.</p><p>There are several enterprise use cases of SLMs. According to a report by <a href="https://www.marketsandmarkets.com/Market-Reports/small-language-model-market-4008452.html" target="_blank"><u>Markets and Markets</u></a>, about 44% of SLM enterprise adoption is driven by the cloud. Enterprises follow a <a href="https://www.itpro.com/hybrid-cloud/29668/what-is-hybrid-cloud">hybrid cloud</a> strategy to lower hardware infrastructure costs.  </p><p>Gartner <a href="https://www.gartner.com/en/newsroom/press-releases/2025-04-09-gartner-predicts-by-2027-organizations-will-use-small-task-specific-ai-models-three-times-more-than-general-purpose-large-language-models" target="_blank">projects</a> that by 2027, organizations will use task-specific SLMs three times more than LLMs.</p><p>SLMs don’t replace LLMs but complement them. The same enterprise may choose SLMs for some secure tasks and LLMs for broader operations. Matt Beucler, CEO and founder of Plura AI, compares LLMs and SLMs.<em> </em>“SLMs handle speed, control, and security", says Beucler. “LLMs do abstraction, synthesis, and creativity. That’s an important conceptual dividing line in modern engineering, and companies that recognize it tend to build scalable, reliable systems.”</p><p>LLMs, Beucler tells <em>ITPro, </em>are also not brilliant at rules-based work on their own. This, and the lure of deploying more industry or case-specific AI models, is part of the appeal of SLMs. In most cases, SLMs are fine-tuned on domain knowledge to perform enterprise-specific tasks. For example, an SLM trained only on medical guidelines and patient data, rather than mathematical equations, would deliver context-aware outputs and exhibit lower chances of hallucinations. </p><p>SLM adoption is also largely driven by training and processing costs. The cost of training SLMs on a smaller dataset is significantly lower than that of LLMs. In 2024, <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-openais-gpt-4o-updates-including-pricing-changes-and-new-features"><u><em>ITPro</em></u><u> reported</u></a> that OpenAI’s GPT-4o mini was 60% cheaper than LLM GPT-3.5 Turbo. In terms of processing costs, LLMs may account for several tens of thousands of dollars across individual tasks, while SLMs reach just a fraction of that.</p><p>The training window of SLMs is as small as a week. LLMs, on the other hand, need months of training to achieve comparable results. SLMs are also less energy-intensive, designed to run locally on low-power devices. As a result, SLMs are an enterprise-favoured choice for <a href="https://www.itpro.com/technology/artificial-intelligence/why-edge-ai-has-significant-business-potential">edge AI</a> and embedded applications. In simple terms, SLMs offer resource efficiency and enhanced cybersecurity at a lower price compared to LLMs.</p><p>However, many SLMs are also derived from a process known as <a href="https://www.itpro.com/technology/artificial-intelligence/google-says-hacker-groups-are-using-gemini-to-augment-attacks-and-companies-are-even-stealing-its-models"><u>‘knowledge distillation’</u></a>, in which a ‘student’ model is trained by a larger model such as the trillion-parameter frontier models. This is important because it shows SLMs aren't a replacement for LLMs altogether.</p><h2 id="use-cases-for-slms">Use cases for SLMs</h2><p>SLM training involves pruning, a model compression technique to remove unnecessary weights from the data. As a result, SLMs fail to answer general and out-of-context queries, making them more useful for specialized applications.</p><p>The developers behind some larger models have released SLMs designed for very specific workloads. For example, in February Alibaba <a href="https://qwen.ai/blog?id=qwen3-coder-next" target="_blank"><u>released</u></a> Qwen-3-Coder-Next, an open-weight model designed for coding agents to run on a single modern Nvidia <a href="https://www.itpro.com/hardware/30399/what-is-a-gpu">GPU</a>.</p><p>Indeed, SLMs are often chosen for code and text generation capabilities. <em>ITPro</em> spoke to Peter Schneider, the senior product manager at Qt Group, a Finnish software company specializing in cross-platform development frameworks, to hear more about practical SLM deployment.</p><p>Schneider shared how they have worked with LLMs and their “student model” SLMs. “We’ve tested code completion on our Qt Modelling Language using several small language models like<strong> </strong><a href="https://www.itpro.com/technology/artificial-intelligence/openai-open-weight-ai-models">GPT-oss-20b</a> or Llama 3.3 70B against larger models like <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform">Gemini 3 Pro</a> Preview, <a href="https://www.itpro.com/technology/artificial-intelligence/openai-thought-it-hit-a-home-run-with-gpt-5-users-werent-so-keen">GPT-5</a>, and <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner">Claude 4.5</a>.”</p><p>“Without tuning, the choice is generally this: do you want a roughly 85% chance of a good code suggestion, or are you ok with about 50%? If you want the former, you go with a frontier model like Gemini or Claude. If you want the latter, then it's a royalty-free small model”.</p><p>Schneider stresses that this doesn’t mean that SLMs should be considered inherently worse options than larger models.</p><p>“That is not to say SLMs can't be good, or even great, for very specific use cases like code completion after you fine-tune the model. We have had success fine-tuning small models like Code Llama to be as good at code completion as bigger models.”</p><p>Defense and aerospace industries deploy SLMs on-prem to handle and analyze classified data. The healthcare industry deploys SLMs for data compliance and protection of patient history. Thanks to the ability of SLMs to function without the internet in case of an emergency. </p><p>“For customers who will never connect to a public cloud in highly regulated industries like medical and defense, the choice is pretty clear”, Schneider highlighted the important enterprise use case of SLMs.</p><p>Searching through data held on premises, or generating documents grounded in information held in a private cloud, is in fact an ideal use case for SLMs. Provided they’re backed with the right hardware, such as a few on premises server racks or on suitably powerful staff devices, they can complete tasks securely and with low latency. </p><p><strong>“</strong>When classifying thousands of support tickets, extracting invoice data, or flagging compliance issues, an SLM trained on your own data will often outperform a general-purpose giant at a fraction of the cost, running in milliseconds and keeping data strictly on-premises,” says Barry James, the director of data and AI at Nucleo.</p><p>Other common SLM use cases include:</p><ul><li><strong>Telecom:</strong> Contact centers are the most common examples where SLMs handle customer data, such as interactive voice response menu (IVR), conversations, ticket routing, FAQs, and product-related queries. With the help of SLMs, enterprises can rely on low-latency volume interactions and use customer data to forecast and predict trends.</li><li><strong>Embedded systems</strong>: SLMs can run without the internet on local edge devices. SLMs are suitable for industrial IoT applications that allow devices to operate in constrained settings with utmost safety features.</li><li><strong>Factories</strong>: SLMs require fewer GPUs to operate. In industries like semiconductor fabrication, SLMs can help build an enterprise-centric knowledge base of common defects and manufacturing errors.</li><li><strong>Banking</strong>: Financial data is one of the most critical types of data in enterprises and banks alike. Regulatory authorities and standards govern bank data globally. The use of AI in such applications is less discussed. With on-premises AI deployment, European banks might be able to comply with the <a href="https://www.itpro.com/it-legislation/27814/what-is-gdpr-everything-you-need-to-know">General Data Protection Regulation (GDPR)</a>.</li></ul><h2 id="notable-slms">Notable SLMs</h2><p>In 2026, some of the most notable SLMs include Microsoft’s Phi-4 and Phi4--mini, Google’s Gemma 3 and Gemma 3n, OpenAI’s gpt-oss-120b and gpt-oss-20b, Alibaba’s Qwen 2.5, and Meta’s Llama 4.</p><p>These models can be run on hardware much smaller than you would find in a data center. Microsoft says Phi-4-mini can be run on a single PC with a minimum of 16GB RAM. Gemma 3n is designed to run on smartphones, while depending on the quantized model size (ranging from 270 million to 27 billion) Gemma 3 can run on a consumer-grade laptop or may require a top-of-the-line GPU.</p><p>As the demand for SLMs increases, AI labs may focus more on quantization and squeezing the most performance possible out of small models.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Researchers taught GPT-5 to learn an obscure programming language on its own ]]></title>
                                                                                                <dc:content><![CDATA[ <p>AI coding tools are surging in popularity, but often require a huge amount of oversight by developers and are trained on specific programming languages. </p><p>But recent research suggests these tools could start filling in gaps in their own knowledge. </p><p>That's according to a <a href="https://arxiv.org/abs/2602.11481" target="_blank"><u>paper</u></a> by researchers at the University of Southern California Viterbi (USC Viterbi), who developed a technique to allow an AI model to improve its own skills in an area it hasn't really been trained on, letting it expand beyond its own training parameters. </p><p>To test the idea, they used OpenAI's GPT-5 and used it to write code in an obscure programming language called Idris – which researchers said they didn't even know how to use themselves. </p><p>Researcher Minda Li improved the model's success from 39% to 96% using a system that allowed her to tell the AI when it was incorrect, but also letting it have another go. </p><p>"Our <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a> are now able to transcend their initial training," said Prof. Bhaskar Krishnamachari in a <a href="https://viterbischool.usc.edu/news/2026/03/the-ai-that-taught-itself-usc-researchers-show-how-artificial-intelligence-can-learn-what-it-never-knew/" target="_blank"><u>blog post</u></a> detailing the project. "Used to be, maybe a year or two ago, you would say an AI model is only as good as the data it has seen. This paper is saying something different."</p><p>The work comes amidst a sharp rise in the use of AI coding tools over the last two years. A <a href="https://www.itpro.com/software/development/ai-coding-is-taking-off-in-the-us-but-developers-in-another-country-are-catching-up-fast" target="_blank"><u>study in </u><u><em>Science</em></u></a><em> </em>from earlier this year<em> </em>showed the share of Python contributions on GitHub that were written by – or otherwise used by – AI in the US climbed from 5% in 2022 to 29% at the end of 2024. </p><h2 id="teaching-gpt-5-to-code-in-idris">Teaching GPT-5 to code in Idris</h2><p>The project made use of Idris, which has just 2,000 code repositories publicly available online versus 24 million for Python – meaning the AI model had much less data to train on. </p><p>Most of the researchers involved on the project had never even heard of it and therefore couldn't tell if the AI's code was correct or not. </p><p>"We were hunting for a language so obscure that we hadn’t heard of it," Krishnamachari said. "I think we were just in my office together, googling around, trying to find some crazy language that no one’s ever heard of." </p><p>That meant Li needed to figure out a way of teaching a skill she couldn't do herself. At first, she set GPT-5 on solving a series of Idris coding exercises on a training platform. It managed a 39% score. </p><p>To boost the score, she gave the model a range of assistance, including access to documentation, manuals, and reference guides – that helped, but only enough for a score around 60%. </p><p>To take it further, Li built a compiler feedback loop, taking error messages from the compiler and giving those to GPT-5 with a prompt to fix the issues. It was given 20 chances per error. </p><p>Li said she expected an improvement of 10%, but instead the system scored nearly perfectly. </p><p>"I was surprised that just that alone, seemingly one simple thing, just keep recompiling, keep trying, was able to get to 96%," she said. </p><h2 id="what-s-next">What's next?</h2><p>The results suggest it's possible to teach AI when there's a lack of available or relevant data, even after the model has been trained. Krishnamachari said the technique could be applied more widely and have huge long-term implications. </p><p>He suggested a feedback loop could help with creating 3D models of buildings or advanced mathematical reasoning, while a former student is working to translate endangered languages. </p><p>However, the technique worked well with coding Idris thanks to the feedback of the compiler, highlighting the continued need for long-standing processes such as <a href="https://www.itpro.com/machine-learning/33704/what-is-reinforcement-learning">reinforcement learning</a> and extensive training. </p><p>"What I’ve learned from this project is that so long as you can figure out how to provide that kind of clear and correct feedback, there’s a chance we can now significantly improve the quality of AI outputs," he noted. </p><p>If that problem can be solved, it should be possible to build "an AI tool to do a task that we cannot do ourselves," Krishnamachari added. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/researchers-taught-openai-gpt-5-to-learn-idris-programming-language-on-its-own</link>
                                                                            <description>
                            <![CDATA[ OpenAI’s GPT-5 learned to code in Idris despite a lack of available data, baffling researchers ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">dQ52tsGDUUBZnQUVRYHCVe</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/PuKbdrZTTMbF2SDsxUAC6D-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Mar 2026 14:00:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:56 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Development]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                    <category><![CDATA[Software]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/PuKbdrZTTMbF2SDsxUAC6D-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Asian female software developer with pink hair working on computer code on a desktop computer in an open plan office space, with laptop screen open on desk.]]></media:description>                                                            <media:text><![CDATA[Asian female software developer with pink hair working on computer code on a desktop computer in an open plan office space, with laptop screen open on desk.]]></media:text>
                                <media:title type="plain"><![CDATA[Asian female software developer with pink hair working on computer code on a desktop computer in an open plan office space, with laptop screen open on desk.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/PuKbdrZTTMbF2SDsxUAC6D-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>AI coding tools are surging in popularity, but often require a huge amount of oversight by developers and are trained on specific programming languages. </p><p>But recent research suggests these tools could start filling in gaps in their own knowledge. </p><p>That's according to a <a href="https://arxiv.org/abs/2602.11481" target="_blank"><u>paper</u></a> by researchers at the University of Southern California Viterbi (USC Viterbi), who developed a technique to allow an AI model to improve its own skills in an area it hasn't really been trained on, letting it expand beyond its own training parameters. </p><p>To test the idea, they used OpenAI's GPT-5 and used it to write code in an obscure programming language called Idris – which researchers said they didn't even know how to use themselves. </p><p>Researcher Minda Li improved the model's success from 39% to 96% using a system that allowed her to tell the AI when it was incorrect, but also letting it have another go. </p><p>"Our <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a> are now able to transcend their initial training," said Prof. Bhaskar Krishnamachari in a <a href="https://viterbischool.usc.edu/news/2026/03/the-ai-that-taught-itself-usc-researchers-show-how-artificial-intelligence-can-learn-what-it-never-knew/" target="_blank"><u>blog post</u></a> detailing the project. "Used to be, maybe a year or two ago, you would say an AI model is only as good as the data it has seen. This paper is saying something different."</p><p>The work comes amidst a sharp rise in the use of AI coding tools over the last two years. A <a href="https://www.itpro.com/software/development/ai-coding-is-taking-off-in-the-us-but-developers-in-another-country-are-catching-up-fast" target="_blank"><u>study in </u><u><em>Science</em></u></a><em> </em>from earlier this year<em> </em>showed the share of Python contributions on GitHub that were written by – or otherwise used by – AI in the US climbed from 5% in 2022 to 29% at the end of 2024. </p><h2 id="teaching-gpt-5-to-code-in-idris">Teaching GPT-5 to code in Idris</h2><p>The project made use of Idris, which has just 2,000 code repositories publicly available online versus 24 million for Python – meaning the AI model had much less data to train on. </p><p>Most of the researchers involved on the project had never even heard of it and therefore couldn't tell if the AI's code was correct or not. </p><p>"We were hunting for a language so obscure that we hadn’t heard of it," Krishnamachari said. "I think we were just in my office together, googling around, trying to find some crazy language that no one’s ever heard of." </p><p>That meant Li needed to figure out a way of teaching a skill she couldn't do herself. At first, she set GPT-5 on solving a series of Idris coding exercises on a training platform. It managed a 39% score. </p><p>To boost the score, she gave the model a range of assistance, including access to documentation, manuals, and reference guides – that helped, but only enough for a score around 60%. </p><p>To take it further, Li built a compiler feedback loop, taking error messages from the compiler and giving those to GPT-5 with a prompt to fix the issues. It was given 20 chances per error. </p><p>Li said she expected an improvement of 10%, but instead the system scored nearly perfectly. </p><p>"I was surprised that just that alone, seemingly one simple thing, just keep recompiling, keep trying, was able to get to 96%," she said. </p><h2 id="what-s-next">What's next?</h2><p>The results suggest it's possible to teach AI when there's a lack of available or relevant data, even after the model has been trained. Krishnamachari said the technique could be applied more widely and have huge long-term implications. </p><p>He suggested a feedback loop could help with creating 3D models of buildings or advanced mathematical reasoning, while a former student is working to translate endangered languages. </p><p>However, the technique worked well with coding Idris thanks to the feedback of the compiler, highlighting the continued need for long-standing processes such as <a href="https://www.itpro.com/machine-learning/33704/what-is-reinforcement-learning">reinforcement learning</a> and extensive training. </p><p>"What I’ve learned from this project is that so long as you can figure out how to provide that kind of clear and correct feedback, there’s a chance we can now significantly improve the quality of AI outputs," he noted. </p><p>If that problem can be solved, it should be possible to build "an AI tool to do a task that we cannot do ourselves," Krishnamachari added. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Microsoft has a new AI poster child in Anthropic ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Microsoft’s cooling relations with OpenAI have been well documented over the last 18 months. While the duo were initially lockstep in leading the generative AI boom, they’ve been slowly drifting apart – and it seems Microsoft has a new poster child in Anthropic. </p><p>With the launch of <a href="https://www.itpro.com/technology/artificial-intelligence/anthropics-claude-cowork-tool-is-coming-to-microsoft-copilot"><u>Cowork for Copilot</u></a>, the integration of Anthropic’s popular Claude Cowork tool with the tech giant’s flagship AI service, Microsoft is finally divesting itself – at least in some capacity – from OpenAI and looking to pastures new. </p><p>This latest development is but one in a number of signs indicating it’s been on the lookout for a new leading partner in the AI space. </p><p>As <em>ITPro </em>previously reported, Microsoft was <a href="https://www.itpro.com/software/microsoft/openai-continues-to-be-our-partner-on-frontier-models-microsoft-is-open-to-using-a-range-of-ai-models-in-365-copilot-but-openai-remains-its-go-to-choice"><u>exploring the possibility of integrating third-party alternatives</u></a> to power Microsoft 365 Copilot as far back as January 2025.</p><p>The tech giant insisted at the time that OpenAI would continue to be a leading partner on frontier models, but in hindsight this was the first major crack in the duo’s happy family veneer. </p><p>September saw Microsoft finally rip off the bandaid in this regard, <a href="https://www.reuters.com/business/microsoft-use-some-ai-anthropic-shift-openai-information-reports-2025-09-09/" target="_blank"><u>signing a deal with Anthropic</u></a> to use its Claude AI models alongside OpenAI’s. </p><h2 id="openai-has-become-a-liability">OpenAI has become a liability</h2><p>It’s hardly surprising Microsoft has been looking elsewhere given the aggravation OpenAI has caused over the last two years. The first clear signs of strain came with the farcical leadership debacle which saw <a href="https://www.itpro.com/technology/artificial-intelligence/openai-chaos-what-next-for-sam-altman-and-the-beleaguered-tech-darling"><u>Sam Altman ousted</u></a> and <a href="https://www.itpro.com/technology/artificial-intelligence/sam-altman-makes-triumphant-return-to-openai-after-days-of-chaos"><u>then return within a matter of days</u></a>. </p><p>That situation was a PR nightmare for Microsoft, and CEO Satya Nadella was <a href="https://www.wired.com/story/openai-sam-altman-ousted-what-happened/" target="_blank"><u>reportedly furious at the events</u></a>. </p><p>That should have been – and very well might have been – the first signs for Microsoft that OpenAI was something of a liability. The company’s activities since then haven’t exactly been reassuring, either. </p><p>As <a href="https://www.itpro.com/technology/artificial-intelligence/2025-marked-the-beginning-of-the-end-for-openai"><u>I’ve mentioned in previous columns for </u><u><em>ITPro</em></u></a>, the company doesn’t appear to have a clear cut idea of what it wants to be. </p><p>Acquisitions aimed at making a charge into the hardware domain highlighted this lack of cohesion and the company’s seemingly endless balancing act catering to both the consumer and enterprise markets has been exhausting. </p><p>And that’s before you get into the array of circular deals and <a href="https://www.itpro.com/business/business-strategy/openai-1-trillion-infrastructure-investment-oracle-broadcom-amd"><u>lavish spending plans</u></a> OpenAI has committed to, which based on <a href="https://www.bloomberg.com/news/articles/2026-03-06/oracle-and-openai-end-plans-to-expand-flagship-data-center"><u>recent reports</u></a> appear to be cracking at the seams. </p><p>Simply put, OpenAI has been so deeply integrated across Microsoft’s core product ranges that if the company even so much as wobbles in its ambitious plans over the next five years, it could spell disaster for the tech giant. Having a backup option – and an <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-reveals-claude-opus-4-6-enterprise-focused-model-1-million-token-context-window"><u>industry leading</u></a> one at that – makes perfect sense. </p><p>Anthropic has proven it’s more than capable of going toe-to-toe with OpenAI over the last two years, consistently <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-promises-opus-level-reasoning-claude-sonnet-4-6-model-at-lower-cost"><u>outperforming its great rival</u></a> on model performance across repeated upgrades. </p><p>Microsoft’s key industry competitors, <a href="https://www.itpro.com/cloud/cloud-computing/aws-european-sovereign-cloud-explained"><u>AWS</u></a> and <a href="https://www.itpro.com/business/public-sector/the-uk-government-is-teaming-up-with-google-cloud-to-kill-public-sector-legacy-tech"><u>Google Cloud</u></a>, both saw the value of Anthropic early on, <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-could-be-the-champion-aws-and-google-need"><u>investing heavily</u></a> in the company in their own attempt to keep pace with the dominance of the Microsoft-OpenAI partnership during the early stages of the generative AI race. </p><h2 id="nobody-wants-a-walled-garden">Nobody wants a walled garden </h2><p>Crucially, Microsoft cozying up to Anthropic makes more sense when one considers the demands of enterprise customers. Nobody wants a return to walled gardens like the early days of the cloud. </p><p>Many businesses are now working on a hybrid or multi-cloud setup, for example, meaning they’ve grown accustomed to flexibility and choice. </p><p>Admittedly, these are two areas in which Microsoft doesn’t have a squeaky clean record. European regulators and trade groups alike would probably have a meltdown seeing “choice”, “flexibility”,  and “Microsoft” uttered in the same sentence. </p><p>Regardless, giving users another option when using its Copilot service adds to the allure for enterprises – and it comes at a crucial time. In the company’s Q2 FY2026 earnings call, <a href="https://www.theregister.com/2026/02/02/microsoft_ai_spend_copilot/"><u>Microsoft hailed record momentum</u></a> with Microsoft 365 Copilot, which it claims now boasts over 15 million paid users. </p><p>Deeper integration of Anthropic models across core products just gives prospective customers more to think about.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/microsoft-has-a-new-ai-poster-child-in-anthropic</link>
                                                                            <description>
                            <![CDATA[ Microsoft is cosying up to Anthropic at a crucial time in the race to deliver on AI promises ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">iqufw4U6uFhKN32qhi572Q</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/jzb4QDynrkMhrvnRfweoJc-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 11 Mar 2026 12:01:46 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:42 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/jzb4QDynrkMhrvnRfweoJc-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A telephoto shot of Satya Nadella, onstage during the keynote of Microsoft Build 2025.]]></media:description>                                                            <media:text><![CDATA[A telephoto shot of Satya Nadella, onstage during the keynote of Microsoft Build 2025.]]></media:text>
                                <media:title type="plain"><![CDATA[A telephoto shot of Satya Nadella, onstage during the keynote of Microsoft Build 2025.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/jzb4QDynrkMhrvnRfweoJc-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Microsoft’s cooling relations with OpenAI have been well documented over the last 18 months. While the duo were initially lockstep in leading the generative AI boom, they’ve been slowly drifting apart – and it seems Microsoft has a new poster child in Anthropic. </p><p>With the launch of <a href="https://www.itpro.com/technology/artificial-intelligence/anthropics-claude-cowork-tool-is-coming-to-microsoft-copilot"><u>Cowork for Copilot</u></a>, the integration of Anthropic’s popular Claude Cowork tool with the tech giant’s flagship AI service, Microsoft is finally divesting itself – at least in some capacity – from OpenAI and looking to pastures new. </p><p>This latest development is but one in a number of signs indicating it’s been on the lookout for a new leading partner in the AI space. </p><p>As <em>ITPro </em>previously reported, Microsoft was <a href="https://www.itpro.com/software/microsoft/openai-continues-to-be-our-partner-on-frontier-models-microsoft-is-open-to-using-a-range-of-ai-models-in-365-copilot-but-openai-remains-its-go-to-choice"><u>exploring the possibility of integrating third-party alternatives</u></a> to power Microsoft 365 Copilot as far back as January 2025.</p><p>The tech giant insisted at the time that OpenAI would continue to be a leading partner on frontier models, but in hindsight this was the first major crack in the duo’s happy family veneer. </p><p>September saw Microsoft finally rip off the bandaid in this regard, <a href="https://www.reuters.com/business/microsoft-use-some-ai-anthropic-shift-openai-information-reports-2025-09-09/" target="_blank"><u>signing a deal with Anthropic</u></a> to use its Claude AI models alongside OpenAI’s. </p><h2 id="openai-has-become-a-liability">OpenAI has become a liability</h2><p>It’s hardly surprising Microsoft has been looking elsewhere given the aggravation OpenAI has caused over the last two years. The first clear signs of strain came with the farcical leadership debacle which saw <a href="https://www.itpro.com/technology/artificial-intelligence/openai-chaos-what-next-for-sam-altman-and-the-beleaguered-tech-darling"><u>Sam Altman ousted</u></a> and <a href="https://www.itpro.com/technology/artificial-intelligence/sam-altman-makes-triumphant-return-to-openai-after-days-of-chaos"><u>then return within a matter of days</u></a>. </p><p>That situation was a PR nightmare for Microsoft, and CEO Satya Nadella was <a href="https://www.wired.com/story/openai-sam-altman-ousted-what-happened/" target="_blank"><u>reportedly furious at the events</u></a>. </p><p>That should have been – and very well might have been – the first signs for Microsoft that OpenAI was something of a liability. The company’s activities since then haven’t exactly been reassuring, either. </p><p>As <a href="https://www.itpro.com/technology/artificial-intelligence/2025-marked-the-beginning-of-the-end-for-openai"><u>I’ve mentioned in previous columns for </u><u><em>ITPro</em></u></a>, the company doesn’t appear to have a clear cut idea of what it wants to be. </p><p>Acquisitions aimed at making a charge into the hardware domain highlighted this lack of cohesion and the company’s seemingly endless balancing act catering to both the consumer and enterprise markets has been exhausting. </p><p>And that’s before you get into the array of circular deals and <a href="https://www.itpro.com/business/business-strategy/openai-1-trillion-infrastructure-investment-oracle-broadcom-amd"><u>lavish spending plans</u></a> OpenAI has committed to, which based on <a href="https://www.bloomberg.com/news/articles/2026-03-06/oracle-and-openai-end-plans-to-expand-flagship-data-center"><u>recent reports</u></a> appear to be cracking at the seams. </p><p>Simply put, OpenAI has been so deeply integrated across Microsoft’s core product ranges that if the company even so much as wobbles in its ambitious plans over the next five years, it could spell disaster for the tech giant. Having a backup option – and an <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-reveals-claude-opus-4-6-enterprise-focused-model-1-million-token-context-window"><u>industry leading</u></a> one at that – makes perfect sense. </p><p>Anthropic has proven it’s more than capable of going toe-to-toe with OpenAI over the last two years, consistently <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-promises-opus-level-reasoning-claude-sonnet-4-6-model-at-lower-cost"><u>outperforming its great rival</u></a> on model performance across repeated upgrades. </p><p>Microsoft’s key industry competitors, <a href="https://www.itpro.com/cloud/cloud-computing/aws-european-sovereign-cloud-explained"><u>AWS</u></a> and <a href="https://www.itpro.com/business/public-sector/the-uk-government-is-teaming-up-with-google-cloud-to-kill-public-sector-legacy-tech"><u>Google Cloud</u></a>, both saw the value of Anthropic early on, <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-could-be-the-champion-aws-and-google-need"><u>investing heavily</u></a> in the company in their own attempt to keep pace with the dominance of the Microsoft-OpenAI partnership during the early stages of the generative AI race. </p><h2 id="nobody-wants-a-walled-garden">Nobody wants a walled garden </h2><p>Crucially, Microsoft cozying up to Anthropic makes more sense when one considers the demands of enterprise customers. Nobody wants a return to walled gardens like the early days of the cloud. </p><p>Many businesses are now working on a hybrid or multi-cloud setup, for example, meaning they’ve grown accustomed to flexibility and choice. </p><p>Admittedly, these are two areas in which Microsoft doesn’t have a squeaky clean record. European regulators and trade groups alike would probably have a meltdown seeing “choice”, “flexibility”,  and “Microsoft” uttered in the same sentence. </p><p>Regardless, giving users another option when using its Copilot service adds to the allure for enterprises – and it comes at a crucial time. In the company’s Q2 FY2026 earnings call, <a href="https://www.theregister.com/2026/02/02/microsoft_ai_spend_copilot/"><u>Microsoft hailed record momentum</u></a> with Microsoft 365 Copilot, which it claims now boasts over 15 million paid users. </p><p>Deeper integration of Anthropic models across core products just gives prospective customers more to think about.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Why Amazon’s ‘go build it’ AI strategy aligns with OpenAI’s big enterprise push ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Amazon’s strategic partnership with OpenAI is a win-win for the duo, according to analysts, as the companies forge closer ties on infrastructure and AI development. </p><p>The pair announced a multi-year deal last week that will see Amazon invest $50 billion in the AI developer, building on a £38 billion pledge made in November last year. </p><p>As part of the deal, OpenAI will lean heavily on AWS infrastructure, committing to deploy 2 gigawatts of Trainium chips for its new enterprise-focused platform, Frontier. AWS will also serve as the exclusive third-party cloud distribution provider for Frontier. </p><p>OpenAI’s Frontier platform allows businesses to build and deploy AI applications and agents. As <em>ITPro </em>reported last month, the platform is a key focus for the firm as it looks to ramp up enterprise adoption rates. </p><p>The company has since brought in <a href="https://www.itpro.com/business/business-strategy/openais-big-enterprise-push-needs-systems-integrators-so-its-turning-to-consultancies-to-plug-implementation-gaps">consultancies to act as systems integrators for customers</a> to streamline adoption processes. </p><p>Elsewhere, the deal will see the duo create a ‘Stateful Runtime Environment’ through <a href="https://www.itpro.com/technology/artificial-intelligence/aws-goes-all-in-on-ai-agents-with-new-features-for-bedrock-and-amazon-q">Amazon Bedrock</a>, again with a focus on development and deployment. This will allow AWS customers to build generative AI apps and agents at “production scale”</p><p>“Stateful developer environments are the next generation of how frontier models will be used, seamlessly enabling models to access elements like compute, memory, and identity,” AWS said in a statement last week. </p><p>“A Stateful Runtime Environment allows developers to keep context, remember prior work, work across software tools and data sources, and access compute.”</p><h2 id="mutually-beneficial-ties">Mutually beneficial ties</h2><p>From Amazon’s perspective, the partnership allows the hyperscaler to continue pushing ahead with its “go build it” approach to AI, according to Lee Sustar, principal analyst at Forrester. </p><p>Amazon’s <a href="https://www.itpro.com/business/business-strategy/how-chief-ai-officers-can-streamline-strategy-from-the-boardroom-down">AI strategy</a> has centered largely around its Amazon Bedrock service, which allows customers to build and deploy AI solutions in a similar vein to OpenAI’s Frontier. It has since built on this with the launch of AgentCore, enabling enterprises to create custom agents. </p><p>With OpenAI’s sharpened focus on enterprise adoption, this represents a mutually beneficial partnership aimed at both pushing broader enterprise uptake while helping AWS capitalize on lucrative infrastructure deals signed by competitors. </p><p>“The OpenAI-AWS deal is the latest example of the hyperscalers’ complicated dance with the big AI players,” Sustar said. “AWS has doubled down on its traditional “go build it” approach with AI, which means it must have partnerships like this.”</p><p>“Notably, it gives AWS exclusive third-party cloud distribution with OpenAI Frontier, potentially giving AWS a competitor to Azure OpenAI Services as well as Google Cloud offerings,” he added. </p><h2 id="diversification-continues">Diversification continues</h2><p>OpenAI’s Trainium commitment marks a huge win for AWS, Sustar noted, allowing the company to “build a rival <a href="https://www.itpro.com/hardware/30399/what-is-a-gpu">GPU </a>ecosystem alternative to Nvidia”.</p><p>Like Google, Amazon has invested heavily in its in-house chip capabilities over the last five years in a bid to reduce reliance on Nvidia hardware. </p><p>The move also highlights OpenAI’s continued efforts to diversify infrastructure and hardware options. </p><p>OpenAI’s once-exclusive ties with Microsoft saw it lean heavily on Azure infrastructure, but over the last year the company has secured a slew of agreements with industry stakeholders on compute capacity and chips. </p><p>The company signed a <a href="https://www.itpro.com/cloud/cloud-computing/oracle-is-reaping-the-rewards-of-openais-huge-compute-requirements">landmark deal with Oracle</a> on compute capacity, for example, while October saw it sign a <a href="https://www.itpro.com/technology/artificial-intelligence/openai-signs-another-chip-deal-this-time-with-amd">multi-billion dollar deal with AMD</a> to use six gigawatts of AMD GPUs.</p><p>In the wake of the Amazon agreement, Microsoft and OpenAI issued a <a href="https://openai.com/index/continuing-microsoft-partnership/"><u>joint statement</u></a> insisting the partnership remains “strong and central”. </p><h2 id="amazon-is-hedging-its-bets">Amazon is hedging its bets</h2><p>While OpenAI is diversifying its infrastructure partners, closer ties with Amazon also highlight the latter’s efforts to broaden its own AI partners.</p><p>Anthropic, for example, initially acted as <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-could-be-the-champion-aws-and-google-need">Amazon’s champion</a> in a similar light to the relationship between OpenAI and Microsoft. </p><p>Amazon, along with Google Cloud, has invested heavily in Anthropic over the last three years through multiple funding pledges. The company’s Claude models are also used extensively by the hyperscaler, powering products such as Alexa+. </p><p>While the deal might have raised questions about the ongoing relationship with Anthropic, Amazon CEO Andy Jassy poured cold water on any speculation last week. </p><p>Speaking to <a href="https://www.cnbc.com/2026/02/27/amazon-open-ai-cloud-jassy-altman.html" target="_blank"><u><em>CNBC</em></u></a>, Jassy said the current deal with OpenAI won’t change relations with the firm's long-time competitor. </p><p>“That relationship will stay strong, and we’re really excited about the partnership we’re building over a long period of time with OpenAI,” he told the broadcaster. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/amazons-go-build-it-ai-strategy-is-a-perfect-fit-for-openais-big-enterprise-push</link>
                                                                            <description>
                            <![CDATA[ OpenAI and Amazon are both vying to offer customers DIY-style AI development services ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">N55iN6rGxCnt8hReWna8Kc</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/gifZ5LVHZ4xfHCCPK3bkPn-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 02 Mar 2026 11:50:34 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:27 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/gifZ5LVHZ4xfHCCPK3bkPn-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Amazon and OpenAI logos pictured against a black backdrop.]]></media:description>                                                            <media:text><![CDATA[Amazon and OpenAI logos pictured against a black backdrop.]]></media:text>
                                <media:title type="plain"><![CDATA[Amazon and OpenAI logos pictured against a black backdrop.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/gifZ5LVHZ4xfHCCPK3bkPn-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Amazon’s strategic partnership with OpenAI is a win-win for the duo, according to analysts, as the companies forge closer ties on infrastructure and AI development. </p><p>The pair announced a multi-year deal last week that will see Amazon invest $50 billion in the AI developer, building on a £38 billion pledge made in November last year. </p><p>As part of the deal, OpenAI will lean heavily on AWS infrastructure, committing to deploy 2 gigawatts of Trainium chips for its new enterprise-focused platform, Frontier. AWS will also serve as the exclusive third-party cloud distribution provider for Frontier. </p><p>OpenAI’s Frontier platform allows businesses to build and deploy AI applications and agents. As <em>ITPro </em>reported last month, the platform is a key focus for the firm as it looks to ramp up enterprise adoption rates. </p><p>The company has since brought in <a href="https://www.itpro.com/business/business-strategy/openais-big-enterprise-push-needs-systems-integrators-so-its-turning-to-consultancies-to-plug-implementation-gaps">consultancies to act as systems integrators for customers</a> to streamline adoption processes. </p><p>Elsewhere, the deal will see the duo create a ‘Stateful Runtime Environment’ through <a href="https://www.itpro.com/technology/artificial-intelligence/aws-goes-all-in-on-ai-agents-with-new-features-for-bedrock-and-amazon-q">Amazon Bedrock</a>, again with a focus on development and deployment. This will allow AWS customers to build generative AI apps and agents at “production scale”</p><p>“Stateful developer environments are the next generation of how frontier models will be used, seamlessly enabling models to access elements like compute, memory, and identity,” AWS said in a statement last week. </p><p>“A Stateful Runtime Environment allows developers to keep context, remember prior work, work across software tools and data sources, and access compute.”</p><h2 id="mutually-beneficial-ties">Mutually beneficial ties</h2><p>From Amazon’s perspective, the partnership allows the hyperscaler to continue pushing ahead with its “go build it” approach to AI, according to Lee Sustar, principal analyst at Forrester. </p><p>Amazon’s <a href="https://www.itpro.com/business/business-strategy/how-chief-ai-officers-can-streamline-strategy-from-the-boardroom-down">AI strategy</a> has centered largely around its Amazon Bedrock service, which allows customers to build and deploy AI solutions in a similar vein to OpenAI’s Frontier. It has since built on this with the launch of AgentCore, enabling enterprises to create custom agents. </p><p>With OpenAI’s sharpened focus on enterprise adoption, this represents a mutually beneficial partnership aimed at both pushing broader enterprise uptake while helping AWS capitalize on lucrative infrastructure deals signed by competitors. </p><p>“The OpenAI-AWS deal is the latest example of the hyperscalers’ complicated dance with the big AI players,” Sustar said. “AWS has doubled down on its traditional “go build it” approach with AI, which means it must have partnerships like this.”</p><p>“Notably, it gives AWS exclusive third-party cloud distribution with OpenAI Frontier, potentially giving AWS a competitor to Azure OpenAI Services as well as Google Cloud offerings,” he added. </p><h2 id="diversification-continues">Diversification continues</h2><p>OpenAI’s Trainium commitment marks a huge win for AWS, Sustar noted, allowing the company to “build a rival <a href="https://www.itpro.com/hardware/30399/what-is-a-gpu">GPU </a>ecosystem alternative to Nvidia”.</p><p>Like Google, Amazon has invested heavily in its in-house chip capabilities over the last five years in a bid to reduce reliance on Nvidia hardware. </p><p>The move also highlights OpenAI’s continued efforts to diversify infrastructure and hardware options. </p><p>OpenAI’s once-exclusive ties with Microsoft saw it lean heavily on Azure infrastructure, but over the last year the company has secured a slew of agreements with industry stakeholders on compute capacity and chips. </p><p>The company signed a <a href="https://www.itpro.com/cloud/cloud-computing/oracle-is-reaping-the-rewards-of-openais-huge-compute-requirements">landmark deal with Oracle</a> on compute capacity, for example, while October saw it sign a <a href="https://www.itpro.com/technology/artificial-intelligence/openai-signs-another-chip-deal-this-time-with-amd">multi-billion dollar deal with AMD</a> to use six gigawatts of AMD GPUs.</p><p>In the wake of the Amazon agreement, Microsoft and OpenAI issued a <a href="https://openai.com/index/continuing-microsoft-partnership/"><u>joint statement</u></a> insisting the partnership remains “strong and central”. </p><h2 id="amazon-is-hedging-its-bets">Amazon is hedging its bets</h2><p>While OpenAI is diversifying its infrastructure partners, closer ties with Amazon also highlight the latter’s efforts to broaden its own AI partners.</p><p>Anthropic, for example, initially acted as <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-could-be-the-champion-aws-and-google-need">Amazon’s champion</a> in a similar light to the relationship between OpenAI and Microsoft. </p><p>Amazon, along with Google Cloud, has invested heavily in Anthropic over the last three years through multiple funding pledges. The company’s Claude models are also used extensively by the hyperscaler, powering products such as Alexa+. </p><p>While the deal might have raised questions about the ongoing relationship with Anthropic, Amazon CEO Andy Jassy poured cold water on any speculation last week. </p><p>Speaking to <a href="https://www.cnbc.com/2026/02/27/amazon-open-ai-cloud-jassy-altman.html" target="_blank"><u><em>CNBC</em></u></a>, Jassy said the current deal with OpenAI won’t change relations with the firm's long-time competitor. </p><p>“That relationship will stay strong, and we’re really excited about the partnership we’re building over a long period of time with OpenAI,” he told the broadcaster. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ OpenAI leans on consultancies to drive enterprise AI adoption ]]></title>
                                                                                                <dc:content><![CDATA[ <p>OpenAI has announced partnerships with a host of big consultancy firms as it looks to ramp up enterprise adoption of its <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a>. </p><p>The firm’s new Frontier Alliances scheme will see it team up with Boston Consulting Group (BCG), McKinsey, Accenture, and Capgemini to help customers “define strategy, integrate systems, redesign workflows, and scale deployment globally”.</p><p>“Our partners will work alongside OpenAI’s Forward Deployed Engineering (FDE) team, combining OpenAI’s research and product expertise with deep transformation experience and global delivery teams,” the company said in a <a href="https://openai.com/index/frontier-alliance-partners/" target="_blank"><u>blog post</u></a>. </p><p>“Each partner is investing in dedicated practice groups and building teams that will be certified on OpenAI technology,” the blog post added. “OpenAI will support them with technical resources, roadmap insight, and access to our product and research teams.”</p><p>The move by OpenAI comes as it looks to bolster enterprise adoption of its AI solutions worldwide. <a href="https://openai.com/index/1-million-businesses-putting-ai-to-work/" target="_blank"><u>Figures published by the firm</u></a> in November 2025 claimed it now boasts over one million business customers, but many enterprises appear to be grappling with implementation challenges.</p><p>OpenAI also <a href="https://openai.com/index/introducing-openai-frontier/" target="_blank"><u>launched its Frontier platform</u></a> in February last year, which aims to help customers build, deploy, and manage agents. </p><p>While this provides the “technical foundation” for enterprises, the firm noted that customer integration still requires support in areas such as strategy planning and broader business transformation practices. </p><p>“The limiting factor for seeing value from AI in enterprises isn’t model intelligence, it’s how agents are built and run in their organizations,” the company said. </p><p>“Making real impact with AI also requires leadership alignment, workflow redesign, integration across systems and data, as well as the kind of change management that drives adoption.”</p><h2 id="openai-needs-systems-integrators">OpenAI needs systems integrators</h2><p>AI isn’t a plug-and-play technology, and businesses across a range of industries have encountered significant challenges with the technology over the last two years.  A study from MIT, for example, found that <a href="https://www.itpro.com/technology/artificial-intelligence/agentic-ai-development-project-delivery-databricks"><u>95% of generative AI pilots fail to reach the production stage</u></a>.</p><p>With agentic AI, the situation isn’t much different for enterprise IT leaders. Recent analysis from Dynatrace found that <a href="https://www.itpro.com/technology/artificial-intelligence/half-of-agentic-ai-projects-are-still-stuck-at-the-pilot-stage-but-thats-not-stopping-enterprises-from-ramping-up-investment"><u>half of agentic AI projects are mired in the proof-of-concept (PoC) stage</u></a>. </p><p>In teaming up with the four consultancy firms, OpenAI hopes to simplify the adoption and roll-out of <a href="https://www.itpro.com/strategy/28181/what-is-ai">AI </a>for customers.</p><p>The company specifically highlighted BCG and McKinsey as boasting “deep expertise” in helping enterprise leaders <a href="https://www.itpro.com/technology/artificial-intelligence/data-strategy-ai-adoption-failures">get started with AI adoption</a>. </p><p>The latter, for example, has a dedicated AI wing, QuantumBlack, which aims to help support businesses across the adoption process. </p><p>“McKinsey brings experience leading enterprise-wide operating change, helping leadership teams align on where to focus, how to redesign operating models, and how to embed intelligence into day-to-day work,” OpenAI said. </p><p>“Through QuantumBlack, its AI arm, McKinsey pairs technical expertise with industry insight to help global clients redesign processes and integrate agents across their highest-value workflows.”</p><p>Accenture and Capgemini, meanwhile, will also act as systems integrators of sorts. </p><p>OpenAI noted that both consultancies will advise on strategy” and help “wire Frontier” into the systems and data enterprises rely on for AI tools. </p><p>Accenture and OpenAI already hold close ties, with the consultancy having equipped thousands of workers with <a href="https://www.itpro.com/technology/artificial-intelligence/chatgpt-enterprise-launches-with-enhanced-data-security-and-record-performance">ChatGPT Enterprise</a>. </p><p>“We’re excited to deepen our work with OpenAI as a Frontier Alliance partner to help clients turn AI into real outcomes,” said Julie Sweet, chair and CEO of Accenture. </p><p>“<a href="https://www.itpro.com/strategy/29899/three-reasons-why-digital-transformation-is-essential-for-business-growth">Business transformation</a> requires more than great models - it requires end-to-end execution across technology, data, security, and change management. Together, we’ll help organizations operationalize AI across the enterprise - responsibly and at scale.”</p><h2 id="ai-is-a-huge-opportunity-for-the-channel">AI is a huge opportunity for the channel</h2><p>The launch of the Frontier Alliances scheme comes amidst a period of huge opportunities for channel partners and systems integrators, who are playing an increasingly vital role in enabling <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> adoption. </p><p><a href="https://omdia.tech.informa.com/blogs/2025/oct/channel-partners-must-evolve-to-lead-genais-transformation" target="_blank"><u>Analysis from Omdia</u></a> in late 2025 found long-running adoption challenges have created a “strategic opportunity” for partners who can plug implementation gaps and help guide enterprises past pilot phases. </p><p>“Partners that build specialized capabilities in production readiness assessment and implementation roadmaps will capture significant market share as organizations struggle to convert promising pilots into enterprise-wide deployments,” the consultancy said. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/business/business-strategy/openais-big-enterprise-push-needs-systems-integrators-so-its-turning-to-consultancies-to-plug-implementation-gaps</link>
                                                                            <description>
                            <![CDATA[ Consultancies such as Accenture and Capgemini will act as systems integrators and help shape AI strategies for OpenAI customers ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">nyn7hNzKMtW87L7Mno5VJ6</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/EFRKmopx9P9fwtUzM6nhpN-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 24 Feb 2026 11:47:06 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 15:00:16 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/EFRKmopx9P9fwtUzM6nhpN-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[OpenAI CEO Sam Altman pictured in a suit with arms folded during an interview in Berlin, Germany.]]></media:description>                                                            <media:text><![CDATA[OpenAI CEO Sam Altman pictured in a suit with arms folded during an interview in Berlin, Germany.]]></media:text>
                                <media:title type="plain"><![CDATA[OpenAI CEO Sam Altman pictured in a suit with arms folded during an interview in Berlin, Germany.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/EFRKmopx9P9fwtUzM6nhpN-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>OpenAI has announced partnerships with a host of big consultancy firms as it looks to ramp up enterprise adoption of its <a href="https://www.itpro.com/technology/artificial-intelligence/amazing-ai-tools-to-try-today">AI tools</a>. </p><p>The firm’s new Frontier Alliances scheme will see it team up with Boston Consulting Group (BCG), McKinsey, Accenture, and Capgemini to help customers “define strategy, integrate systems, redesign workflows, and scale deployment globally”.</p><p>“Our partners will work alongside OpenAI’s Forward Deployed Engineering (FDE) team, combining OpenAI’s research and product expertise with deep transformation experience and global delivery teams,” the company said in a <a href="https://openai.com/index/frontier-alliance-partners/" target="_blank"><u>blog post</u></a>. </p><p>“Each partner is investing in dedicated practice groups and building teams that will be certified on OpenAI technology,” the blog post added. “OpenAI will support them with technical resources, roadmap insight, and access to our product and research teams.”</p><p>The move by OpenAI comes as it looks to bolster enterprise adoption of its AI solutions worldwide. <a href="https://openai.com/index/1-million-businesses-putting-ai-to-work/" target="_blank"><u>Figures published by the firm</u></a> in November 2025 claimed it now boasts over one million business customers, but many enterprises appear to be grappling with implementation challenges.</p><p>OpenAI also <a href="https://openai.com/index/introducing-openai-frontier/" target="_blank"><u>launched its Frontier platform</u></a> in February last year, which aims to help customers build, deploy, and manage agents. </p><p>While this provides the “technical foundation” for enterprises, the firm noted that customer integration still requires support in areas such as strategy planning and broader business transformation practices. </p><p>“The limiting factor for seeing value from AI in enterprises isn’t model intelligence, it’s how agents are built and run in their organizations,” the company said. </p><p>“Making real impact with AI also requires leadership alignment, workflow redesign, integration across systems and data, as well as the kind of change management that drives adoption.”</p><h2 id="openai-needs-systems-integrators">OpenAI needs systems integrators</h2><p>AI isn’t a plug-and-play technology, and businesses across a range of industries have encountered significant challenges with the technology over the last two years.  A study from MIT, for example, found that <a href="https://www.itpro.com/technology/artificial-intelligence/agentic-ai-development-project-delivery-databricks"><u>95% of generative AI pilots fail to reach the production stage</u></a>.</p><p>With agentic AI, the situation isn’t much different for enterprise IT leaders. Recent analysis from Dynatrace found that <a href="https://www.itpro.com/technology/artificial-intelligence/half-of-agentic-ai-projects-are-still-stuck-at-the-pilot-stage-but-thats-not-stopping-enterprises-from-ramping-up-investment"><u>half of agentic AI projects are mired in the proof-of-concept (PoC) stage</u></a>. </p><p>In teaming up with the four consultancy firms, OpenAI hopes to simplify the adoption and roll-out of <a href="https://www.itpro.com/strategy/28181/what-is-ai">AI </a>for customers.</p><p>The company specifically highlighted BCG and McKinsey as boasting “deep expertise” in helping enterprise leaders <a href="https://www.itpro.com/technology/artificial-intelligence/data-strategy-ai-adoption-failures">get started with AI adoption</a>. </p><p>The latter, for example, has a dedicated AI wing, QuantumBlack, which aims to help support businesses across the adoption process. </p><p>“McKinsey brings experience leading enterprise-wide operating change, helping leadership teams align on where to focus, how to redesign operating models, and how to embed intelligence into day-to-day work,” OpenAI said. </p><p>“Through QuantumBlack, its AI arm, McKinsey pairs technical expertise with industry insight to help global clients redesign processes and integrate agents across their highest-value workflows.”</p><p>Accenture and Capgemini, meanwhile, will also act as systems integrators of sorts. </p><p>OpenAI noted that both consultancies will advise on strategy” and help “wire Frontier” into the systems and data enterprises rely on for AI tools. </p><p>Accenture and OpenAI already hold close ties, with the consultancy having equipped thousands of workers with <a href="https://www.itpro.com/technology/artificial-intelligence/chatgpt-enterprise-launches-with-enhanced-data-security-and-record-performance">ChatGPT Enterprise</a>. </p><p>“We’re excited to deepen our work with OpenAI as a Frontier Alliance partner to help clients turn AI into real outcomes,” said Julie Sweet, chair and CEO of Accenture. </p><p>“<a href="https://www.itpro.com/strategy/29899/three-reasons-why-digital-transformation-is-essential-for-business-growth">Business transformation</a> requires more than great models - it requires end-to-end execution across technology, data, security, and change management. Together, we’ll help organizations operationalize AI across the enterprise - responsibly and at scale.”</p><h2 id="ai-is-a-huge-opportunity-for-the-channel">AI is a huge opportunity for the channel</h2><p>The launch of the Frontier Alliances scheme comes amidst a period of huge opportunities for channel partners and systems integrators, who are playing an increasingly vital role in enabling <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> adoption. </p><p><a href="https://omdia.tech.informa.com/blogs/2025/oct/channel-partners-must-evolve-to-lead-genais-transformation" target="_blank"><u>Analysis from Omdia</u></a> in late 2025 found long-running adoption challenges have created a “strategic opportunity” for partners who can plug implementation gaps and help guide enterprises past pilot phases. </p><p>“Partners that build specialized capabilities in production readiness assessment and implementation roadmaps will capture significant market share as organizations struggle to convert promising pilots into enterprise-wide deployments,” the consultancy said. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Microsoft Copilot bug saw AI snoop on confidential emails — after it was told not to ]]></title>
                                                                                                <dc:content><![CDATA[ <p><a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-copilot-review-ai-baked-into-your-apps">Microsoft's Copilot</a> has been found reading and summarizing email messages despite "confidential" labels that should prevent the AI system from accessing the data. </p><p>The tech giant issued a warning about a bug in the Microsoft 365 Copilot "work tab" Chat which allows the AI to incorrectly process messages that should be skipped due to sensitivity labels. </p><p>In a <a href="https://support.nhs.net/2026/02/microsoft-365-alert-service-degradation-power-bi-users-pipelines-associated-with-dataflow-gen2-refreshes-in-microsoft-fabric-show-activity-status-as-failed-2/" target="_blank"><u>message</u></a> shared to affected users, Microsoft said a code issue meant emails in the sent items and draft folders were being picked up despite policies in place that meant messages with confidential labels shouldn't be read. </p><p>"We identified and addressed an issue where Microsoft 365 Copilot Chat could return content from emails labeled confidential authored by a user and stored within their Draft and Sent Items in Outlook desktop," a spokesperson told <em>ITPro</em>.</p><p>"This did not provide anyone access to information they weren’t already authorized to see. While our access controls and data protection policies remained intact, this behavior did not meet our intended Copilot experience, which is designed to exclude protected content from Copilot access".</p><p>The spokesperson added that a "configuration update" has been deployed for customers globally. </p><p>The issue was first spotted on 21 January, and tracked by Microsoft as CW1226324. </p><h2 id="microsoft-copilot-chat-rules">Microsoft Copilot Chat rules </h2><p>Copilot Chat is Microsoft's tool for interacting with an AI agent directly from Word and other <a href="https://www.itpro.com/software/development/microsoft-claims-ai-is-augmenting-developers-rather-than-replacing-them">productivity software</a>. It first rolled out in September. </p><p>Microsoft 365 Copilot reads through data such as emails, documents, chats, and more to help dig information out for users. </p><p>With privacy in mind, Microsoft built in administrative controls that let companies keep AI away from sensitive material — but this bug meant those rules were not applied in <em>Sent Items </em>and <em>Drafts </em>folders in email, letting Copilot access all emails for summarization despite being labelled confidential. </p><h2 id="ai-security-risks">AI security risks</h2><p>The rise of generative AI use in businesses has sparked concerns about the security risks, be it breaching confidentiality guidelines in sensitive industries, leaking private data, or offering a new attack vector via <a href="https://www.itpro.com/technology/artificial-intelligence/openai-chatgpt-atlas-ai-browser-prompt-injection-attack-risk"><u>prompt injections</u></a> or other hacking techniques.  </p><p>Researchers have <a href="https://www.itpro.com/security/12-000-api-keys-and-passwords-were-found-in-a-popular-ai-training-dataset-experts-say-the-issue-is-down-to-poor-identity-management"><u>already spotted thousands of corporate secrets</u></a> in one popular AI training dataset, suggesting industry is struggling to keep up with the realities of data security in the AI era. </p><p>The risk is exacerbated by <a href="https://www.itpro.com/technology/artificial-intelligence/gartner-says-40-percent-of-enterprises-will-experience-shadow-ai-breaches-by-2030-educating-staff-is-the-key-to-avoiding-disaster"><u>shadow AI</u></a>, when employees use AI chatbots or other tools without official approval or IT department support, meaning data-protection guidelines aren't in place to protect private or sensitive information. </p><p>That's already causing a <a href="https://www.itpro.com/technology/artificial-intelligence/generative-ai-data-violations-more-than-doubled-last-year"><u>huge surge in data policy violations</u></a>, according to a report from Netskope, with almost a <a href="https://www.itpro.com/security/data-protection/almost-a-third-of-workers-are-covertly-using-ai-at-work-heres-what-thats-a-terrible-idea"><u>third of workers already using AI covertly at work</u></a>.</p><p>There have been previous issues with Copilot. Back in 2024, <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-copilot-could-have-serious-vulnerabilities-after-researchers-reveal-data-leak-issues-in-rag-systems"><u>academic researchers spotted security vulnerabilities</u></a> in retrieval augmented generation (RAG) systems used by Microsoft Copilot that could lead to such tools committing confidentiality violations.</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/microsoft-copilot-bug-saw-ai-snoop-on-confidential-emails-after-it-was-told-not-to</link>
                                                                            <description>
                            <![CDATA[ The Copilot bug meant an AI summarizing tool accessed messages in the Sent and Draft folders, dodging policy rules ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">DLTvd6EVzR2BHccEnfZbmj</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/6ii6y9BQKbHMTYapsbBtfg-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 19 Feb 2026 15:44:50 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:17 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/6ii6y9BQKbHMTYapsbBtfg-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Microsoft 365 Copilot branding pictured on smartphone screen held in hand.]]></media:description>                                                            <media:text><![CDATA[Microsoft 365 Copilot branding pictured on smartphone screen held in hand.]]></media:text>
                                <media:title type="plain"><![CDATA[Microsoft 365 Copilot branding pictured on smartphone screen held in hand.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/6ii6y9BQKbHMTYapsbBtfg-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p><a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-copilot-review-ai-baked-into-your-apps">Microsoft's Copilot</a> has been found reading and summarizing email messages despite "confidential" labels that should prevent the AI system from accessing the data. </p><p>The tech giant issued a warning about a bug in the Microsoft 365 Copilot "work tab" Chat which allows the AI to incorrectly process messages that should be skipped due to sensitivity labels. </p><p>In a <a href="https://support.nhs.net/2026/02/microsoft-365-alert-service-degradation-power-bi-users-pipelines-associated-with-dataflow-gen2-refreshes-in-microsoft-fabric-show-activity-status-as-failed-2/" target="_blank"><u>message</u></a> shared to affected users, Microsoft said a code issue meant emails in the sent items and draft folders were being picked up despite policies in place that meant messages with confidential labels shouldn't be read. </p><p>"We identified and addressed an issue where Microsoft 365 Copilot Chat could return content from emails labeled confidential authored by a user and stored within their Draft and Sent Items in Outlook desktop," a spokesperson told <em>ITPro</em>.</p><p>"This did not provide anyone access to information they weren’t already authorized to see. While our access controls and data protection policies remained intact, this behavior did not meet our intended Copilot experience, which is designed to exclude protected content from Copilot access".</p><p>The spokesperson added that a "configuration update" has been deployed for customers globally. </p><p>The issue was first spotted on 21 January, and tracked by Microsoft as CW1226324. </p><h2 id="microsoft-copilot-chat-rules">Microsoft Copilot Chat rules </h2><p>Copilot Chat is Microsoft's tool for interacting with an AI agent directly from Word and other <a href="https://www.itpro.com/software/development/microsoft-claims-ai-is-augmenting-developers-rather-than-replacing-them">productivity software</a>. It first rolled out in September. </p><p>Microsoft 365 Copilot reads through data such as emails, documents, chats, and more to help dig information out for users. </p><p>With privacy in mind, Microsoft built in administrative controls that let companies keep AI away from sensitive material — but this bug meant those rules were not applied in <em>Sent Items </em>and <em>Drafts </em>folders in email, letting Copilot access all emails for summarization despite being labelled confidential. </p><h2 id="ai-security-risks">AI security risks</h2><p>The rise of generative AI use in businesses has sparked concerns about the security risks, be it breaching confidentiality guidelines in sensitive industries, leaking private data, or offering a new attack vector via <a href="https://www.itpro.com/technology/artificial-intelligence/openai-chatgpt-atlas-ai-browser-prompt-injection-attack-risk"><u>prompt injections</u></a> or other hacking techniques.  </p><p>Researchers have <a href="https://www.itpro.com/security/12-000-api-keys-and-passwords-were-found-in-a-popular-ai-training-dataset-experts-say-the-issue-is-down-to-poor-identity-management"><u>already spotted thousands of corporate secrets</u></a> in one popular AI training dataset, suggesting industry is struggling to keep up with the realities of data security in the AI era. </p><p>The risk is exacerbated by <a href="https://www.itpro.com/technology/artificial-intelligence/gartner-says-40-percent-of-enterprises-will-experience-shadow-ai-breaches-by-2030-educating-staff-is-the-key-to-avoiding-disaster"><u>shadow AI</u></a>, when employees use AI chatbots or other tools without official approval or IT department support, meaning data-protection guidelines aren't in place to protect private or sensitive information. </p><p>That's already causing a <a href="https://www.itpro.com/technology/artificial-intelligence/generative-ai-data-violations-more-than-doubled-last-year"><u>huge surge in data policy violations</u></a>, according to a report from Netskope, with almost a <a href="https://www.itpro.com/security/data-protection/almost-a-third-of-workers-are-covertly-using-ai-at-work-heres-what-thats-a-terrible-idea"><u>third of workers already using AI covertly at work</u></a>.</p><p>There have been previous issues with Copilot. Back in 2024, <a href="https://www.itpro.com/technology/artificial-intelligence/microsoft-copilot-could-have-serious-vulnerabilities-after-researchers-reveal-data-leak-issues-in-rag-systems"><u>academic researchers spotted security vulnerabilities</u></a> in retrieval augmented generation (RAG) systems used by Microsoft Copilot that could lead to such tools committing confidentiality violations.</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Anthropic promises ‘Opus-level’ reasoning with new Claude Sonnet 4.6 model ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Anthropic has announced Claude Sonnet 4.6, its latest ‘all rounder’ AI model which it says can achieve results previously only possible with its most sophisticated Opus family of models – and all at a fraction of the price.</p><p>Claude Sonnet 4.6 is intended as an extremely capable model for coding and computer use, with tangible performance improvements across real-world, economically-valuable tasks.</p><p>Anthropic stressed that these new capabilities are paired with API costs low enough for businesses to reasonably use it as a backend model for high-performance AI agents, or to upgrade the outputs of their existing Claude Sonnet deployments at no extra cost.</p><p>Indeed, via the Claude API, the new model is accessible at the same price as Claude Sonnet 4.5: $3 per million input tokens and $15 per million output tokens.</p><p>This compares to the Claude Opus 4.6 pricing of $5 per million input tokens and $25 per million output tokens.</p><p>But industry experts describe its performance as significantly improved compared to the previous Sonnet release.</p><p>"The performance-to-cost ratio of Claude Sonnet 4.6 is extraordinary—it’s hard to overstate how fast Claude models have been evolving in recent months,” said Michele Catasta, president at <a href="https://www.itpro.com/technology/artificial-intelligence/the-replit-vibe-coding-incident-gives-us-a-glimpse-into-why-developers-are-still-wary-of-ai-coding-assistants"><u>Replit</u></a>.</p><p>“Sonnet 4.6 outperforms on our orchestration evals, handles our most complex agentic workloads, and keeps improving the higher you push the effort settings."</p><h2 id="claude-sonnet-4-6-benchmarks-how-does-it-compare">Claude Sonnet 4.6 benchmarks: How does it compare?</h2><p>In the popular coding benchmark SWE-bench Verified, Claude Sonnet 4.6 scored 79.6%, comparable to the score of other frontier models.</p><p>Perhaps more importantly, real-world testing found developers in <a href="https://www.itpro.com/software/development/claude-code-creator-boris-cherny-says-software-engineers-are-more-important-than-ever-as-ai-transforms-the-profession-but-anthropic-ceo-dario-amodei-still-thinks-full-automation-is-coming"><u>Claude Code</u></a> preferred Claude Sonnet 4.6 outputs around 70% of the time – and even found its code outputs more thorough than Claude Opus 4.5, which until recently had been Anthropic’s most powerful model.</p><p>The new model also boasts a one million token context window, in beta, enough to hold entire code bases or multiple complex documents for context.</p><p>Claude Sonnet 4.6 also demonstrates a significant leap in agentic tool use and computer use compared to its predecessor. </p><p>For example in OSWorld-Verified, a benchmark intended to test how well AI agents can use web and desktop apps, Claude Sonnet 4.6 scored 72.5%, a marked improvement on the 61.4% scored by the previous Sonnet model and even more than the 66.3% achieved by Claude Opus 4.5.</p><p>Crucially, Claude Sonnet 4.6’s score in this benchmark is only 0.2% behind that of Anthropic’s flagship model <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-reveals-claude-opus-4-6-enterprise-focused-model-1-million-token-context-window"><u>Claude Opus 4.6, launched two weeks ago</u></a>.</p><p>In the real-world office task benchmark GDBval-AA-Elo, Claude Sonnet 4.6 even outperforms Claude Opus 4.6, a feat it also achieves in the agentic financial analysis benchmark Finance Agent v1.1.</p><p>"Claude Sonnet 4.6 matches Opus 4.6 performance on OfficeQA, which measures how well a model can read enterprise documents (charts, PDFs, tables), pull the right facts, and reason from those facts,” said Hanling Tang, CTO of Neural Networks at Databricks. </p><p>“It’s a meaningful upgrade for document comprehension workloads."</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:1548px;"><p class="vanilla-image-block" style="padding-top:115.25%;"><img id="nJDFRdienw7p9HjbksJnKL" name="Claude Sonnet 4.6 benchmarks" alt="A benchmark table for Claude Sonnet 4.6, compared to Opus 4.5, Sonnet 4.5, Gemini 3 Pro, and GPT-5.2. Claude Opus 4.6 performs similarly to the other frontier models at almost every benchmark, and beats them at GDBval-AA-Elo and Finance Agent v1.1." src="https://cdn.mos.cms.futurecdn.net/nJDFRdienw7p9HjbksJnKL-1920-80.jpg" mos="" align="middle" fullscreen="" width="1548" height="1784" attribution="" endorsement="" class="inline" /></p></div></div><figcaption itemprop="caption description" class=" inline-layout"><span class="credit" itemprop="copyrightHolder">(Image credit: Anthropic)</span></figcaption></figure><h2 id="how-to-access-claude-sonnet-4-6">How to access Claude Sonnet 4.6</h2><p>Claude Sonnet 4.6 is accessible now via all channels to access Anthropic’s models, including the Claude API, <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-anthropic-claude-cowork"><u>Claude Cowork</u></a>, Claude subscription plans, and via public cloud platforms. </p><p>In its announcement, Anthropic confirmed that Claude Sonnet 4.6 is now the default model on the Claude web app, as well as Claude Cowork, and has upgraded its free-tier capabilities to include file creation, <a href="https://www.itpro.com/business/want-to-get-the-most-out-of-anthropics-claude-ai-assistant-this-new-training-course-will-give-you-prompt-engineering-tips-and-how-to-use-claude-code"><u>skills</u></a>, and code compaction to optimize long conversations.</p><p>The new model launches alongside new support for MCP connectors inside Claude in Excel, the company revealed.</p><p>This allows users to pull external context into Claude from within their spreadsheet and to connect directly to tools such as Daloopa, FactSet, LSEG, Moody’s, Pitchbook, and S&amp;P Global.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/anthropic-promises-opus-level-reasoning-claude-sonnet-4-6-model-at-lower-cost</link>
                                                                            <description>
                            <![CDATA[ The latest addition to the Claude family is explicitly intended to power AI agents, with pricing and capabilities designed to attract enterprise attention ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">MTE8mm7VqA79vTskACqT6W</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/DXRJWrNgbLPhiFnGkYVoDV-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Feb 2026 18:00:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:48 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ rory.bathgate@futurenet.com (Rory Bathgate) ]]></author>                    <dc:creator><![CDATA[ Rory Bathgate ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Vep5JogbPhduK7R6CUWAm6-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Rory Bathgate was the Features and Multimedia Editor at ITPro until 2026, overseeing all in-depth content and case studies. Throughout his time at ITPro, Rory  charted the rise in popularity of generative AI and specifically companies such as Microsoft, OpenAI, and Google. &lt;/p&gt;&lt;p&gt;Alongside this, he delved into increasing calls for ethical and responsible AI as global legislators circle the technology, as well as the latest in mobile networking technology, from 5G mmWave to the 3G sunset and how it will affect businesses.&lt;/p&gt;&lt;p&gt;He  provided coverage from high-profile tech conferences such as Dell Technologies World, SuiteWorld, and VMware Explore Europe. His on-the-ground coverage  included live blogs, extensive daily coverage of the most significant announcements, analysis pieces, and podcast.&lt;/p&gt;&lt;p&gt;Rory was the co-host of the ITPro Podcast alongside Jane McCallion, where he swapped a keyboard for a microphone to discuss the latest learnings in tech.&lt;/p&gt;&lt;p&gt;Rory has an MA in Eighteenth-Century Studies from King’s College London, as well as a BA in English and American Literature from the University of Kent. He joined ITPro in 2022 as a graduate, after four years in student journalism.&lt;/p&gt;&lt;p&gt;In his free time, Rory enjoys photography and video editing, and can often be found at the cinema or reading a good science fiction paperback.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/DXRJWrNgbLPhiFnGkYVoDV-1920-80.jpg">
                                                            <media:credit><![CDATA[Anthropic]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[The logo for Anthropic&#039;s Claude AI model on a cream background.]]></media:description>                                                            <media:text><![CDATA[The logo for Anthropic&#039;s Claude AI model on a cream background.]]></media:text>
                                <media:title type="plain"><![CDATA[The logo for Anthropic&#039;s Claude AI model on a cream background.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/DXRJWrNgbLPhiFnGkYVoDV-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Anthropic has announced Claude Sonnet 4.6, its latest ‘all rounder’ AI model which it says can achieve results previously only possible with its most sophisticated Opus family of models – and all at a fraction of the price.</p><p>Claude Sonnet 4.6 is intended as an extremely capable model for coding and computer use, with tangible performance improvements across real-world, economically-valuable tasks.</p><p>Anthropic stressed that these new capabilities are paired with API costs low enough for businesses to reasonably use it as a backend model for high-performance AI agents, or to upgrade the outputs of their existing Claude Sonnet deployments at no extra cost.</p><p>Indeed, via the Claude API, the new model is accessible at the same price as Claude Sonnet 4.5: $3 per million input tokens and $15 per million output tokens.</p><p>This compares to the Claude Opus 4.6 pricing of $5 per million input tokens and $25 per million output tokens.</p><p>But industry experts describe its performance as significantly improved compared to the previous Sonnet release.</p><p>"The performance-to-cost ratio of Claude Sonnet 4.6 is extraordinary—it’s hard to overstate how fast Claude models have been evolving in recent months,” said Michele Catasta, president at <a href="https://www.itpro.com/technology/artificial-intelligence/the-replit-vibe-coding-incident-gives-us-a-glimpse-into-why-developers-are-still-wary-of-ai-coding-assistants"><u>Replit</u></a>.</p><p>“Sonnet 4.6 outperforms on our orchestration evals, handles our most complex agentic workloads, and keeps improving the higher you push the effort settings."</p><h2 id="claude-sonnet-4-6-benchmarks-how-does-it-compare">Claude Sonnet 4.6 benchmarks: How does it compare?</h2><p>In the popular coding benchmark SWE-bench Verified, Claude Sonnet 4.6 scored 79.6%, comparable to the score of other frontier models.</p><p>Perhaps more importantly, real-world testing found developers in <a href="https://www.itpro.com/software/development/claude-code-creator-boris-cherny-says-software-engineers-are-more-important-than-ever-as-ai-transforms-the-profession-but-anthropic-ceo-dario-amodei-still-thinks-full-automation-is-coming"><u>Claude Code</u></a> preferred Claude Sonnet 4.6 outputs around 70% of the time – and even found its code outputs more thorough than Claude Opus 4.5, which until recently had been Anthropic’s most powerful model.</p><p>The new model also boasts a one million token context window, in beta, enough to hold entire code bases or multiple complex documents for context.</p><p>Claude Sonnet 4.6 also demonstrates a significant leap in agentic tool use and computer use compared to its predecessor. </p><p>For example in OSWorld-Verified, a benchmark intended to test how well AI agents can use web and desktop apps, Claude Sonnet 4.6 scored 72.5%, a marked improvement on the 61.4% scored by the previous Sonnet model and even more than the 66.3% achieved by Claude Opus 4.5.</p><p>Crucially, Claude Sonnet 4.6’s score in this benchmark is only 0.2% behind that of Anthropic’s flagship model <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-reveals-claude-opus-4-6-enterprise-focused-model-1-million-token-context-window"><u>Claude Opus 4.6, launched two weeks ago</u></a>.</p><p>In the real-world office task benchmark GDBval-AA-Elo, Claude Sonnet 4.6 even outperforms Claude Opus 4.6, a feat it also achieves in the agentic financial analysis benchmark Finance Agent v1.1.</p><p>"Claude Sonnet 4.6 matches Opus 4.6 performance on OfficeQA, which measures how well a model can read enterprise documents (charts, PDFs, tables), pull the right facts, and reason from those facts,” said Hanling Tang, CTO of Neural Networks at Databricks. </p><p>“It’s a meaningful upgrade for document comprehension workloads."</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:1548px;"><p class="vanilla-image-block" style="padding-top:115.25%;"><img id="nJDFRdienw7p9HjbksJnKL" name="Claude Sonnet 4.6 benchmarks" alt="A benchmark table for Claude Sonnet 4.6, compared to Opus 4.5, Sonnet 4.5, Gemini 3 Pro, and GPT-5.2. Claude Opus 4.6 performs similarly to the other frontier models at almost every benchmark, and beats them at GDBval-AA-Elo and Finance Agent v1.1." src="https://cdn.mos.cms.futurecdn.net/nJDFRdienw7p9HjbksJnKL-1920-80.jpg" mos="" align="middle" fullscreen="" width="1548" height="1784" attribution="" endorsement="" class="inline" /></p></div></div><figcaption itemprop="caption description" class=" inline-layout"><span class="credit" itemprop="copyrightHolder">(Image credit: Anthropic)</span></figcaption></figure><h2 id="how-to-access-claude-sonnet-4-6">How to access Claude Sonnet 4.6</h2><p>Claude Sonnet 4.6 is accessible now via all channels to access Anthropic’s models, including the Claude API, <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-anthropic-claude-cowork"><u>Claude Cowork</u></a>, Claude subscription plans, and via public cloud platforms. </p><p>In its announcement, Anthropic confirmed that Claude Sonnet 4.6 is now the default model on the Claude web app, as well as Claude Cowork, and has upgraded its free-tier capabilities to include file creation, <a href="https://www.itpro.com/business/want-to-get-the-most-out-of-anthropics-claude-ai-assistant-this-new-training-course-will-give-you-prompt-engineering-tips-and-how-to-use-claude-code"><u>skills</u></a>, and code compaction to optimize long conversations.</p><p>The new model launches alongside new support for MCP connectors inside Claude in Excel, the company revealed.</p><p>This allows users to pull external context into Claude from within their spreadsheet and to connect directly to tools such as Daloopa, FactSet, LSEG, Moody’s, Pitchbook, and S&amp;P Global.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Anthropic reveals Claude Opus 4.6, its new enterprise-focused model ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Anthropic has announced Claude Opus 4.6, a major upgrade to its flagship AI model focused on knowledge work and agentic coding tasks.</p><p>The model is intended to complete long, complex enterprise tasks in a single shot with higher-quality outputs across text, presentations, and spreadsheets than previous models released by Anthropic.</p><p>Code generation is a core focus for Claude Opus 4.6, which boasts a one million token context window – in beta – capable of using entire code bases for reference when generating code outputs.</p><p>Anthropic’s models are well-regarded for their raw capabilities at code generation, with <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner"><u>Claude Opus 4.5</u></a> having set a new bar across coding benchmarks upon release in November 2025.</p><p>Claude Opus 4.6 achieves industry-leading scores across a number of popular benchmarks, including 65.4% in the coding benchmark Terminal-Bench 2.0 and 1,606 at GDPval-AA Elo, which tests performance for economically-valuable knowledge work.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:1536px;"><p class="vanilla-image-block" style="padding-top:114.32%;"><img id="P8uQJLdKUZCAFbNTTUbKqQ" name="Claude Opus 4.6 benchmarks" alt="A benchmark table for Claude Opus 4.6, compared to Opus 4.5, Sonnet 4.5, Gemini 3 Pro, and GPT-5.2. Claude Opus 4.6 outperforms the other models at almost every benchmark." src="https://cdn.mos.cms.futurecdn.net/P8uQJLdKUZCAFbNTTUbKqQ-1920-80.jpg" mos="" align="middle" fullscreen="" width="1536" height="1756" attribution="" endorsement="" class="inline" /></p></div></div><figcaption itemprop="caption description" class=" inline-layout"><span class="credit" itemprop="copyrightHolder">(Image credit: Anthropic)</span></figcaption></figure><p>Overall, the new model is also several points ahead of all other frontier AI models across enterprise-specific benchmarks including sector-specific tests for financial and legal capabilities.</p><p>Alongside Claude Opus 4.6, developers will also be able to use a new ‘Agent Teams’ capability as a research preview in <a href="https://www.itpro.com/software/development/anthropic-claude-code-for-web-closed-beta-launch">Claude Code</a>. </p><p>This allows multiple agents powered by Claude Opus 4.6 to work together in parallel, coordinating on a shared goal and reducing agentic processing bottlenecks.</p><h2 id="api-controls-and-better-knowledge-retrieval">API controls and better knowledge retrieval</h2><p>Leveraging its large context window, Claude Opus 4.6 is intended to be far more useful for accessing information held in large datasets as well as long-form coding tasks such as bug detection and more complicated code implementation.</p><p>While competitor models such as Gemini 3 – as well as Anthropic’s own Claude Sonnet family of models – have boasted similar context windows in the past, this is a first for the Claude Opus family.</p><p>Anthropic said it has improved Claude’s ability to retain performance even when presented with a huge number of input tokens. </p><p>Tested with the ‘needle-in-a-haystack’ benchmark MRCR v2, which tests model fact-finding, attention, and reasoning across long and complex prompts, Claude Opus 4.6 scored 76% compared with Claude Sonnet 4.5’s 18.5%.</p><p>Alongside the new model, Anthropic announced additional controls for developers on the Claude API.</p><p>New controls include ‘adaptive thinking’, a configurable value that sets the degree to which Claude Opus 4.6 will use context clues to decide when its extended thinking capabilities are needed, as well as four ‘effort’ levels: low, medium, high (the default) and max.</p><p>Using this control, developers can reduce latency and inference costs when using the model for less complex problems or to power agents tasked with relatively simple workloads.</p><p>Long conversations in the Claude API will also now make use of context compaction, an offering in beta, which summarizes older conversational tokens to free up room in the context window over long back-and-forth tasks. </p><p>Context compaction happens when conversations hit a specific length, which developers can control.</p><h2 id="enterprise-breakthroughs">Enterprise breakthroughs</h2><p>Claude Opus 4.6 is intended as major leap for enterprise AI, with the largest improvements over its predecessor found in knowledge retrieval and sector-specific tool use.</p><p>In part, Anthropic is achieving this through new integrations for Claude directly within Microsoft 365, with improved data processing for Claude for Excel and new native slideshow support via Claude in PowerPoint.</p><p>In a demo video, Anthropic showed how Claude Opus 4.6 could be used to ingest enterprise spreadsheets and produce detailed competitor analysis, outputting new spreadsheets and an entire PowerPoint deck containing the most pertinent information.</p><p>Customer feedback on the new model emphasizes its strength across a wide range of complex enterprise tasks including codebase migration, agentic tool use, and extensive knowledge work.</p><p>“Claude Opus 4.6 handled a multi-million-line codebase migration like a senior engineer,” said Gregor Stewart, chief AI officer at SentinelOne.</p><p>“It planned upfront, adapted its strategy as it learned, and finished in half the time.”</p><p>Customers also praised the model’s extended context, particularly for knowledge work such as handling legal documents.</p><p>“Anthropic’s latest model represents a meaningful leap in long-context performance,” said Joel Hron, CTO at Thomson Reuters.</p><p>“In our testing, we saw it handle much larger bodies of information with a level of consistency that strengthens how we design and deploy complex research workflows. Progress in this area gives us more powerful building blocks to deliver truly expert-grade systems professionals can trust.”</p><p>Earlier this week, the legal giant suffered an 18% slump in its stocks in a selloff many investors <a href="https://www.itpro.com/technology/artificial-intelligence/why-anthropic-sent-software-stocks-into-freefall"><u>linked to the launch of Anthropic’s enterprise-specific tool Cowork</u></a> which can be used to automate some legal work.</p><p>“Claude Opus 4.6 excels in high-reasoning tasks, like multi-source analysis, across legal, financial, and technical content,” said Yashodha Bhavnani, head of AI at Box.  </p><p>“Box’s eval showed a 10% lift in performance, reaching 68% vs. a 58% baseline, and near-perfect scores in technical domains.”</p><p>In practical tests, Claude Opus 4.6 also showed strong capabilities in managing IT issues as an autonomous agent.</p><p>“Claude Opus 4.6 autonomously closed 13 issues and assigned 12 issues to the right team members in a single day, managing a ~50-person organization across 6 repositories,” said Yusuke Kaji, general manager, AI at Rakuten. </p><p>“It handled both product and organizational decisions while synthesizing context across multiple domains, and knew when to escalate to a human.”</p><p>Claude Opus 4.6 is available at the same price as Claude Opus 4.5 via Anthropic’s API: $5 per million input tokens and $25 per million output tokens.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/anthropic-reveals-claude-opus-4-6-enterprise-focused-model-1-million-token-context-window</link>
                                                                            <description>
                            <![CDATA[ The AI developer highlighted financial and legal tasks, as well as agent tool use, as particular strengths for the new model ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">uST7A5ubMZComGKrxDWhT3</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/mgDQATGQbZpP7s9dSEskag-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 05 Feb 2026 17:45:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:48 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ rory.bathgate@futurenet.com (Rory Bathgate) ]]></author>                    <dc:creator><![CDATA[ Rory Bathgate ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Vep5JogbPhduK7R6CUWAm6-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Rory Bathgate was the Features and Multimedia Editor at ITPro until 2026, overseeing all in-depth content and case studies. Throughout his time at ITPro, Rory  charted the rise in popularity of generative AI and specifically companies such as Microsoft, OpenAI, and Google. &lt;/p&gt;&lt;p&gt;Alongside this, he delved into increasing calls for ethical and responsible AI as global legislators circle the technology, as well as the latest in mobile networking technology, from 5G mmWave to the 3G sunset and how it will affect businesses.&lt;/p&gt;&lt;p&gt;He  provided coverage from high-profile tech conferences such as Dell Technologies World, SuiteWorld, and VMware Explore Europe. His on-the-ground coverage  included live blogs, extensive daily coverage of the most significant announcements, analysis pieces, and podcast.&lt;/p&gt;&lt;p&gt;Rory was the co-host of the ITPro Podcast alongside Jane McCallion, where he swapped a keyboard for a microphone to discuss the latest learnings in tech.&lt;/p&gt;&lt;p&gt;Rory has an MA in Eighteenth-Century Studies from King’s College London, as well as a BA in English and American Literature from the University of Kent. He joined ITPro in 2022 as a graduate, after four years in student journalism.&lt;/p&gt;&lt;p&gt;In his free time, Rory enjoys photography and video editing, and can often be found at the cinema or reading a good science fiction paperback.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/mgDQATGQbZpP7s9dSEskag-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A photo of the Anthropic Claude logo on a phone in someone&#039;s hands, with the word &#039;Claude&#039; and logo against a white backdrop visible in the background.]]></media:description>                                                            <media:text><![CDATA[A photo of the Anthropic Claude logo on a phone in someone&#039;s hands, with the word &#039;Claude&#039; and logo against a white backdrop visible in the background.]]></media:text>
                                <media:title type="plain"><![CDATA[A photo of the Anthropic Claude logo on a phone in someone&#039;s hands, with the word &#039;Claude&#039; and logo against a white backdrop visible in the background.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/mgDQATGQbZpP7s9dSEskag-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Anthropic has announced Claude Opus 4.6, a major upgrade to its flagship AI model focused on knowledge work and agentic coding tasks.</p><p>The model is intended to complete long, complex enterprise tasks in a single shot with higher-quality outputs across text, presentations, and spreadsheets than previous models released by Anthropic.</p><p>Code generation is a core focus for Claude Opus 4.6, which boasts a one million token context window – in beta – capable of using entire code bases for reference when generating code outputs.</p><p>Anthropic’s models are well-regarded for their raw capabilities at code generation, with <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner"><u>Claude Opus 4.5</u></a> having set a new bar across coding benchmarks upon release in November 2025.</p><p>Claude Opus 4.6 achieves industry-leading scores across a number of popular benchmarks, including 65.4% in the coding benchmark Terminal-Bench 2.0 and 1,606 at GDPval-AA Elo, which tests performance for economically-valuable knowledge work.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:1536px;"><p class="vanilla-image-block" style="padding-top:114.32%;"><img id="P8uQJLdKUZCAFbNTTUbKqQ" name="Claude Opus 4.6 benchmarks" alt="A benchmark table for Claude Opus 4.6, compared to Opus 4.5, Sonnet 4.5, Gemini 3 Pro, and GPT-5.2. Claude Opus 4.6 outperforms the other models at almost every benchmark." src="https://cdn.mos.cms.futurecdn.net/P8uQJLdKUZCAFbNTTUbKqQ-1920-80.jpg" mos="" align="middle" fullscreen="" width="1536" height="1756" attribution="" endorsement="" class="inline" /></p></div></div><figcaption itemprop="caption description" class=" inline-layout"><span class="credit" itemprop="copyrightHolder">(Image credit: Anthropic)</span></figcaption></figure><p>Overall, the new model is also several points ahead of all other frontier AI models across enterprise-specific benchmarks including sector-specific tests for financial and legal capabilities.</p><p>Alongside Claude Opus 4.6, developers will also be able to use a new ‘Agent Teams’ capability as a research preview in <a href="https://www.itpro.com/software/development/anthropic-claude-code-for-web-closed-beta-launch">Claude Code</a>. </p><p>This allows multiple agents powered by Claude Opus 4.6 to work together in parallel, coordinating on a shared goal and reducing agentic processing bottlenecks.</p><h2 id="api-controls-and-better-knowledge-retrieval">API controls and better knowledge retrieval</h2><p>Leveraging its large context window, Claude Opus 4.6 is intended to be far more useful for accessing information held in large datasets as well as long-form coding tasks such as bug detection and more complicated code implementation.</p><p>While competitor models such as Gemini 3 – as well as Anthropic’s own Claude Sonnet family of models – have boasted similar context windows in the past, this is a first for the Claude Opus family.</p><p>Anthropic said it has improved Claude’s ability to retain performance even when presented with a huge number of input tokens. </p><p>Tested with the ‘needle-in-a-haystack’ benchmark MRCR v2, which tests model fact-finding, attention, and reasoning across long and complex prompts, Claude Opus 4.6 scored 76% compared with Claude Sonnet 4.5’s 18.5%.</p><p>Alongside the new model, Anthropic announced additional controls for developers on the Claude API.</p><p>New controls include ‘adaptive thinking’, a configurable value that sets the degree to which Claude Opus 4.6 will use context clues to decide when its extended thinking capabilities are needed, as well as four ‘effort’ levels: low, medium, high (the default) and max.</p><p>Using this control, developers can reduce latency and inference costs when using the model for less complex problems or to power agents tasked with relatively simple workloads.</p><p>Long conversations in the Claude API will also now make use of context compaction, an offering in beta, which summarizes older conversational tokens to free up room in the context window over long back-and-forth tasks. </p><p>Context compaction happens when conversations hit a specific length, which developers can control.</p><h2 id="enterprise-breakthroughs">Enterprise breakthroughs</h2><p>Claude Opus 4.6 is intended as major leap for enterprise AI, with the largest improvements over its predecessor found in knowledge retrieval and sector-specific tool use.</p><p>In part, Anthropic is achieving this through new integrations for Claude directly within Microsoft 365, with improved data processing for Claude for Excel and new native slideshow support via Claude in PowerPoint.</p><p>In a demo video, Anthropic showed how Claude Opus 4.6 could be used to ingest enterprise spreadsheets and produce detailed competitor analysis, outputting new spreadsheets and an entire PowerPoint deck containing the most pertinent information.</p><p>Customer feedback on the new model emphasizes its strength across a wide range of complex enterprise tasks including codebase migration, agentic tool use, and extensive knowledge work.</p><p>“Claude Opus 4.6 handled a multi-million-line codebase migration like a senior engineer,” said Gregor Stewart, chief AI officer at SentinelOne.</p><p>“It planned upfront, adapted its strategy as it learned, and finished in half the time.”</p><p>Customers also praised the model’s extended context, particularly for knowledge work such as handling legal documents.</p><p>“Anthropic’s latest model represents a meaningful leap in long-context performance,” said Joel Hron, CTO at Thomson Reuters.</p><p>“In our testing, we saw it handle much larger bodies of information with a level of consistency that strengthens how we design and deploy complex research workflows. Progress in this area gives us more powerful building blocks to deliver truly expert-grade systems professionals can trust.”</p><p>Earlier this week, the legal giant suffered an 18% slump in its stocks in a selloff many investors <a href="https://www.itpro.com/technology/artificial-intelligence/why-anthropic-sent-software-stocks-into-freefall"><u>linked to the launch of Anthropic’s enterprise-specific tool Cowork</u></a> which can be used to automate some legal work.</p><p>“Claude Opus 4.6 excels in high-reasoning tasks, like multi-source analysis, across legal, financial, and technical content,” said Yashodha Bhavnani, head of AI at Box.  </p><p>“Box’s eval showed a 10% lift in performance, reaching 68% vs. a 58% baseline, and near-perfect scores in technical domains.”</p><p>In practical tests, Claude Opus 4.6 also showed strong capabilities in managing IT issues as an autonomous agent.</p><p>“Claude Opus 4.6 autonomously closed 13 issues and assigned 12 issues to the right team members in a single day, managing a ~50-person organization across 6 repositories,” said Yusuke Kaji, general manager, AI at Rakuten. </p><p>“It handled both product and organizational decisions while synthesizing context across multiple domains, and knew when to escalate to a human.”</p><p>Claude Opus 4.6 is available at the same price as Claude Opus 4.5 via Anthropic’s API: $5 per million input tokens and $25 per million output tokens.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Sundar Pichai hails AI gains as Google Cloud, Gemini growth surges ]]></title>
                                                                                                <dc:content><![CDATA[ <p><a href="https://www.itpro.com/software/development/google-ceo-sundar-pichai-says-vibe-coding-has-made-software-development-exciting-again-developers-might-disagree">Google CEO Sundar Pichai</a> has hailed the company’s growing strength in the AI race after its cloud unit recorded significant financial growth. </p><p>Alphabet’s 2025 financial year ended on a bang, with revenue surpassing $400 billion for the first time and Google Cloud revenue reaching $17.7 billion – a 48% increase compared to the same period last year. </p><p>Pichai claimed growth in this division is a clear signal of the company’s ongoing success in <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a>. </p><p>"Overall, we’re seeing our AI investments and infrastructure drive revenue and growth across the board," he told investors.</p><p>Google Cloud has been the <a href="https://www.itpro.com/technology/artificial-intelligence/google-cloud-next-2025-enterprise-ai-adoption">tip of the spear for the tech giant in the AI race</a>, leading on the roll-out of its flagship Gemini model range and development of in-house hardware for <a href="https://www.itpro.com/infrastructure/the-role-of-the-cpu-in-the-ai-era">AI inference</a>.</p><p>With Gemini, the growth signals are clear. Pichai revealed Gemini 3 Pro has recorded the “fastest adoption of any model in our history”. </p><p>Indeed, the AI app now boasts more than 750 million active monthly users, marking an increase from 650 million in the previous quarter. Pichai also told investors the company secured eight million paid seats for Gemini Enterprise packages in the space of four months. </p><h2 id="google-cloud-is-gaining-ground-albeit-slowly">Google Cloud is gaining ground, albeit slowly</h2><p>Lee Sustar, principal analyst at Forrester, said Google Cloud’s revenue growth is a key indicator that Alphabet is gaining ground on cloud competitors such as Amazon Web Services (AWS) and Microsoft.</p><p>“Google Cloud's quarterly revenue jump of 48% over the same period a year earlier is decisive evidence that it is a full-blown enterprise challenger to AWS and Microsoft Azure,” he said. </p><p>The hyperscaler still lags behind AWS and Microsoft both in terms of market share and revenue, however. </p><p><a href="https://www.srgresearch.com/articles/cloud-market-share-trends-big-three-together-hold-63-while-oracle-and-the-neoclouds-inch-higher" target="_blank"><u>Figures from Statista</u></a> show AWS retains a dominant position in the cloud computing market with a 29% share, but both Microsoft and Google Cloud have been <a href="https://www.itpro.com/cloud/cloud-computing/is-aws-cloud-dominance-waning-new-stats-show-the-hyperscalers-iaas-market-share-is-decreasing-while-microsoft-and-google-record-gains"><u>gaining ground in recent years</u></a>. </p><p>Microsoft, Google Cloud’s closest competitor, still dwarfs it in terms of revenue. The tech giant’s <a href="https://www.microsoft.com/en-us/investor/earnings/fy-2026-q2/press-release-webcast" target="_blank"><u>recent earnings report</u></a> shows its cloud unit recorded 26% growth across the quarter and surpassed $50 billion in revenue - nearly three-times that of Google Cloud. </p><h2 id="bullish-spending-plans">Bullish spending plans </h2><p>While Alphabet earnings highlight the continued success of its AI and infrastructure push, Sustar noted that growth “comes at a hefty price” as capital expenditure projections soared. </p><p>The company plans to spend between $175 billion and $185 billion across 2026, roughly twice what it spent last year and a significant increase on analyst expectations of $115 billion. </p><p>Pichai attributed this to supply constraints and future planning, with the company appearing keen to strike while the iron is hot with bullish infrastructure investment.</p><p>“We’ve been supply constrained even as we’ve been ramping up our capacity,” he told investors. “Obviously, our capex spend this year is an eye towards the future. We are constantly planning for the long-term.”</p><p>Much of this spending will go toward compute capacity to meet the aforementioned constraints, according to CFO Anat Ashkenazi, with around 60% of spending focused on upgrading existing hardware assets and construction of new data center infrastructure.</p><h2 id=""></h2><p>Google was famously caught off guard by the launch of ChatGPT in late 2022 and had a rocky start following a lackluster product launch with Bard in February 2023.</p><p>However, company appears to have cast off initial concerns about its ability to compete with Microsoft and OpenAI.</p><p>The <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform">launch of Gemini 3 Pro</a> in November 2025 marked a significant turning point for the tech giant. The new flagship model vastly outperformed competitor options such as GPT-5.1 and Claude Sonnet 4.5 on multimodal reasoning benchmarks, for example. </p><p>MMMU-Pro testing showed Gemini 3 Pro scored 81% versus GPT-5.1’s 76% and Claude Sonnet 4.5’s 68%.</p><p>Hardware is another area in which Google has recorded positive signs. At Google Cloud Next 2025 the company <a href="https://www.itpro.com/infrastructure/google-cloud-ironwood-chip-hypercomputer-upgrades"><u>unveiled ‘Ironwood’</u></a>, its seventh generation TPU that, at the time, offered industry-leading capabilities for AI workloads.</p><p>While Google still uses Nvidia hardware across its infrastructure, this has become a notable USP for the company, giving it a degree of flexibility in meeting customer compute demands as well as a buffer against chip supply constraints.</p><p>As <em>ITPro </em>reported last year, the <a href="https://www.itpro.com/infrastructure/why-google-cloud-is-betting-big-on-its-custom-chips">tech giant has invested heavily in its TPU range</a>, and previous models including Gemini 2.5 Pro and Gemma 3 were trained using the hardware. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/the-fastest-adoption-of-any-model-in-our-history-sundar-pichai-hails-ai-gains-as-google-cloud-growth-gemini-popularity-surges</link>
                                                                            <description>
                            <![CDATA[ The company’s cloud unit beat Wall Street expectations as it continues to play a key role in driving AI adoption ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">vmiddghRGukKJQSsNrmWuG</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/ZT5kQmajbHnWdnqwnLZ9BW-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 05 Feb 2026 12:32:53 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:39 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/ZT5kQmajbHnWdnqwnLZ9BW-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Google CEO Sundar Pichai pictured speaking on stage at the New York Times annual DealBook summit.]]></media:description>                                                            <media:text><![CDATA[Google CEO Sundar Pichai pictured speaking on stage at the New York Times annual DealBook summit.]]></media:text>
                                <media:title type="plain"><![CDATA[Google CEO Sundar Pichai pictured speaking on stage at the New York Times annual DealBook summit.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/ZT5kQmajbHnWdnqwnLZ9BW-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p><a href="https://www.itpro.com/software/development/google-ceo-sundar-pichai-says-vibe-coding-has-made-software-development-exciting-again-developers-might-disagree">Google CEO Sundar Pichai</a> has hailed the company’s growing strength in the AI race after its cloud unit recorded significant financial growth. </p><p>Alphabet’s 2025 financial year ended on a bang, with revenue surpassing $400 billion for the first time and Google Cloud revenue reaching $17.7 billion – a 48% increase compared to the same period last year. </p><p>Pichai claimed growth in this division is a clear signal of the company’s ongoing success in <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a>. </p><p>"Overall, we’re seeing our AI investments and infrastructure drive revenue and growth across the board," he told investors.</p><p>Google Cloud has been the <a href="https://www.itpro.com/technology/artificial-intelligence/google-cloud-next-2025-enterprise-ai-adoption">tip of the spear for the tech giant in the AI race</a>, leading on the roll-out of its flagship Gemini model range and development of in-house hardware for <a href="https://www.itpro.com/infrastructure/the-role-of-the-cpu-in-the-ai-era">AI inference</a>.</p><p>With Gemini, the growth signals are clear. Pichai revealed Gemini 3 Pro has recorded the “fastest adoption of any model in our history”. </p><p>Indeed, the AI app now boasts more than 750 million active monthly users, marking an increase from 650 million in the previous quarter. Pichai also told investors the company secured eight million paid seats for Gemini Enterprise packages in the space of four months. </p><h2 id="google-cloud-is-gaining-ground-albeit-slowly">Google Cloud is gaining ground, albeit slowly</h2><p>Lee Sustar, principal analyst at Forrester, said Google Cloud’s revenue growth is a key indicator that Alphabet is gaining ground on cloud competitors such as Amazon Web Services (AWS) and Microsoft.</p><p>“Google Cloud's quarterly revenue jump of 48% over the same period a year earlier is decisive evidence that it is a full-blown enterprise challenger to AWS and Microsoft Azure,” he said. </p><p>The hyperscaler still lags behind AWS and Microsoft both in terms of market share and revenue, however. </p><p><a href="https://www.srgresearch.com/articles/cloud-market-share-trends-big-three-together-hold-63-while-oracle-and-the-neoclouds-inch-higher" target="_blank"><u>Figures from Statista</u></a> show AWS retains a dominant position in the cloud computing market with a 29% share, but both Microsoft and Google Cloud have been <a href="https://www.itpro.com/cloud/cloud-computing/is-aws-cloud-dominance-waning-new-stats-show-the-hyperscalers-iaas-market-share-is-decreasing-while-microsoft-and-google-record-gains"><u>gaining ground in recent years</u></a>. </p><p>Microsoft, Google Cloud’s closest competitor, still dwarfs it in terms of revenue. The tech giant’s <a href="https://www.microsoft.com/en-us/investor/earnings/fy-2026-q2/press-release-webcast" target="_blank"><u>recent earnings report</u></a> shows its cloud unit recorded 26% growth across the quarter and surpassed $50 billion in revenue - nearly three-times that of Google Cloud. </p><h2 id="bullish-spending-plans">Bullish spending plans </h2><p>While Alphabet earnings highlight the continued success of its AI and infrastructure push, Sustar noted that growth “comes at a hefty price” as capital expenditure projections soared. </p><p>The company plans to spend between $175 billion and $185 billion across 2026, roughly twice what it spent last year and a significant increase on analyst expectations of $115 billion. </p><p>Pichai attributed this to supply constraints and future planning, with the company appearing keen to strike while the iron is hot with bullish infrastructure investment.</p><p>“We’ve been supply constrained even as we’ve been ramping up our capacity,” he told investors. “Obviously, our capex spend this year is an eye towards the future. We are constantly planning for the long-term.”</p><p>Much of this spending will go toward compute capacity to meet the aforementioned constraints, according to CFO Anat Ashkenazi, with around 60% of spending focused on upgrading existing hardware assets and construction of new data center infrastructure.</p><h2 id=""></h2><p>Google was famously caught off guard by the launch of ChatGPT in late 2022 and had a rocky start following a lackluster product launch with Bard in February 2023.</p><p>However, company appears to have cast off initial concerns about its ability to compete with Microsoft and OpenAI.</p><p>The <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform">launch of Gemini 3 Pro</a> in November 2025 marked a significant turning point for the tech giant. The new flagship model vastly outperformed competitor options such as GPT-5.1 and Claude Sonnet 4.5 on multimodal reasoning benchmarks, for example. </p><p>MMMU-Pro testing showed Gemini 3 Pro scored 81% versus GPT-5.1’s 76% and Claude Sonnet 4.5’s 68%.</p><p>Hardware is another area in which Google has recorded positive signs. At Google Cloud Next 2025 the company <a href="https://www.itpro.com/infrastructure/google-cloud-ironwood-chip-hypercomputer-upgrades"><u>unveiled ‘Ironwood’</u></a>, its seventh generation TPU that, at the time, offered industry-leading capabilities for AI workloads.</p><p>While Google still uses Nvidia hardware across its infrastructure, this has become a notable USP for the company, giving it a degree of flexibility in meeting customer compute demands as well as a buffer against chip supply constraints.</p><p>As <em>ITPro </em>reported last year, the <a href="https://www.itpro.com/infrastructure/why-google-cloud-is-betting-big-on-its-custom-chips">tech giant has invested heavily in its TPU range</a>, and previous models including Gemini 2.5 Pro and Gemma 3 were trained using the hardware. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Amazon's OpenAI investment points to troubled waters for the hyperscaler ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Amazon’s rumored OpenAI investment plans have drawn criticism from analysts amid suggestions the move shows a “lack of confidence” in its own in-house AI model range. </p><p>According to reports from the <a href="https://www.wsj.com/tech/ai/amazon-in-talks-to-invest-up-to-50-billion-in-openai-43191ba0?gaa_at=eafs&amp;gaa_n=AWEtsqe_DDMvXjHYDhgPG-JxC0w6XvB6ZkSzq2mfy6zlE3TpHqgH2ObK9_qrWgMKM84%3D&amp;gaa_ts=69808139&amp;gaa_sig=aq6ro4ZunroD_FGlsVzOIQdZyJgfcMEQWmb55oD1e4NEkoWx6IYjH6MjvniVgqyKRfiOmZ_VTmPsIPM7hXtxAQ%3D%3D" target="_blank"><u><em>Wall Street Journal</em></u></a>, the hyperscaler could invest up to $50 billion in the AI company as part of a multi-stage funding round which includes backing from Nvidia, Microsoft, and SoftBank. </p><p>While exact details on the funding remain unclear, OpenAI has been seeking suitors for an additional $100 billion in investment, largely to fund future infrastructure projects. </p><p>Amazon boasts close ties with Anthropic, a key competitor for OpenAI in the <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> space. <a href="https://www.itpro.com/cloud/infrastructure-as-a-service-iaas/362608/what-is-aws">Amazon Web Services (AWS)</a> has served as the company’s primary cloud provider since 2023, and in 2024 signed a fresh agreement to become the firm’s primary training and inference partner. </p><p>In recent months, reports suggest a warming of ties between Amazon and OpenAI, particularly as the latter embarked on a spending spree. </p><p><a href="https://www.cnbc.com/2025/12/16/openai-in-talks-with-amazon-about-investment-could-top-10-billion.html" target="_blank"><u><em>CNBC</em></u></a> reported in December that the duo were locked in investment discussions, with OpenAI apparently keen on an agreement which would allow access to Amazon’s in-house AI chips. </p><h2 id="what-the-openai-investment-suggests">What the OpenAI investment suggests</h2><p>OpenAI has made no secret about its fundraising efforts over the last two years. In November 2024, the firm <a href="https://www.itpro.com/technology/artificial-intelligence/openai-just-raised-usd6-6bn-in-funding-but-it-drove-a-hard-bargain-investors-will-be-stopped-from-funding-rivals"><u>raised $6.6 billion in funding</u></a> at a $157 billion valuation, but it hasn’t stopped there. </p><p>2025 was a whirlwind year, with the AI developer signing a host of megadeals with industry titans such as Oracle, AMD, <a href="https://www.itpro.com/technology/artificial-intelligence/nvidia-openai-investment-100-billion"><u>Nvidia</u></a>, and Broadcom. </p><p>For OpenAI, the aims are crystal clear: it has an insatiable appetite for compute power and it’s burning cash at an alarming rate. </p><p>The company has lined up more than $1 trillion in spending over the next decade, prompting repeated concerns about its long-term viability and whether it can meet the terms of deals signed with Oracle, for example. </p><p>With this in mind, backing from Amazon – and crucially, investment – would mark a major seal of approval for the company. From Amazon’s perspective, the situation appears rather convoluted, according to Forrester principal analyst Rowan Curran. </p><p>For Amazon to justify this investment, the deal will likely have to center around heavy use of AWS infrastructure for training and inferencing purposes, as well as use of Amazon’s in-house chips. </p><p>“Amazon’s reported coming investment in OpenAI would be a tremendous expenditure of capital that will only benefit Amazon if that money goes right into OpenAI using AWS servers for training and inferencing,” he said. </p><p>From an AI model perspective, the deal doesn’t seem to add up for the hyperscaler, which already hosts OpenAI models through its Bedrock service.</p><p>Enterprises are increasingly using a mixture of options rather than sticking with a monolithic approach to the technology – a trend that AWS sought to capitalize on with the launch of Bedrock in 2023. </p><p>“The foundation model market is not commoditized, but investments of this scale into a single company do not reflect the realities of an increasingly diverse model market, and the demands of mature customers who are using an array of models to achieve their goals, rather than a single model,” Curran commented. </p><p>Similarly, the company has invested heavily in its own in-house Nova AI model range. </p><p>Amazon unveiled the launch of the Nova models at the 2025 AWS re:Invent conference in Las Vegas. As <a href="https://www.itpro.com/cloud/live/aws-re-invent-2025-all-the-news-updates-and-announcements-live-from-las-vegas"><em>ITPro </em>reported in December</a>, these models have since received an array of updates and improvements.</p><p>Notably, however, Curran suggested this could signal a “lack of confidence” in the Nova models and perhaps a growing sense of frustration at progress on this front. </p><p>“This external investment would further signal a lack of confidence in AWS's internal family of Nova models to achieve a level of functionality and usage with customers that will produce significant ROI,” he noted. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/amazons-rumored-openai-investment-points-to-a-lack-of-confidence-in-nova-model-range</link>
                                                                            <description>
                            <![CDATA[ The hyperscaler is among a number of firms targeting investment in the company ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">WVCHUKyggeXtt4e23z99mZ</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/2YmEYkDwiABjNXPnG4ESwc-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 02 Feb 2026 12:00:31 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:06 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/2YmEYkDwiABjNXPnG4ESwc-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Amazon CEO Andy Jassy speaks during a keynote address at AWS re:Invent 2024 in Las Vegas, Nevada, USA.]]></media:description>                                                            <media:text><![CDATA[Amazon CEO Andy Jassy speaks during a keynote address at AWS re:Invent 2024 in Las Vegas, Nevada, USA.]]></media:text>
                                <media:title type="plain"><![CDATA[Amazon CEO Andy Jassy speaks during a keynote address at AWS re:Invent 2024 in Las Vegas, Nevada, USA.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/2YmEYkDwiABjNXPnG4ESwc-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Amazon’s rumored OpenAI investment plans have drawn criticism from analysts amid suggestions the move shows a “lack of confidence” in its own in-house AI model range. </p><p>According to reports from the <a href="https://www.wsj.com/tech/ai/amazon-in-talks-to-invest-up-to-50-billion-in-openai-43191ba0?gaa_at=eafs&amp;gaa_n=AWEtsqe_DDMvXjHYDhgPG-JxC0w6XvB6ZkSzq2mfy6zlE3TpHqgH2ObK9_qrWgMKM84%3D&amp;gaa_ts=69808139&amp;gaa_sig=aq6ro4ZunroD_FGlsVzOIQdZyJgfcMEQWmb55oD1e4NEkoWx6IYjH6MjvniVgqyKRfiOmZ_VTmPsIPM7hXtxAQ%3D%3D" target="_blank"><u><em>Wall Street Journal</em></u></a>, the hyperscaler could invest up to $50 billion in the AI company as part of a multi-stage funding round which includes backing from Nvidia, Microsoft, and SoftBank. </p><p>While exact details on the funding remain unclear, OpenAI has been seeking suitors for an additional $100 billion in investment, largely to fund future infrastructure projects. </p><p>Amazon boasts close ties with Anthropic, a key competitor for OpenAI in the <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">generative AI</a> space. <a href="https://www.itpro.com/cloud/infrastructure-as-a-service-iaas/362608/what-is-aws">Amazon Web Services (AWS)</a> has served as the company’s primary cloud provider since 2023, and in 2024 signed a fresh agreement to become the firm’s primary training and inference partner. </p><p>In recent months, reports suggest a warming of ties between Amazon and OpenAI, particularly as the latter embarked on a spending spree. </p><p><a href="https://www.cnbc.com/2025/12/16/openai-in-talks-with-amazon-about-investment-could-top-10-billion.html" target="_blank"><u><em>CNBC</em></u></a> reported in December that the duo were locked in investment discussions, with OpenAI apparently keen on an agreement which would allow access to Amazon’s in-house AI chips. </p><h2 id="what-the-openai-investment-suggests">What the OpenAI investment suggests</h2><p>OpenAI has made no secret about its fundraising efforts over the last two years. In November 2024, the firm <a href="https://www.itpro.com/technology/artificial-intelligence/openai-just-raised-usd6-6bn-in-funding-but-it-drove-a-hard-bargain-investors-will-be-stopped-from-funding-rivals"><u>raised $6.6 billion in funding</u></a> at a $157 billion valuation, but it hasn’t stopped there. </p><p>2025 was a whirlwind year, with the AI developer signing a host of megadeals with industry titans such as Oracle, AMD, <a href="https://www.itpro.com/technology/artificial-intelligence/nvidia-openai-investment-100-billion"><u>Nvidia</u></a>, and Broadcom. </p><p>For OpenAI, the aims are crystal clear: it has an insatiable appetite for compute power and it’s burning cash at an alarming rate. </p><p>The company has lined up more than $1 trillion in spending over the next decade, prompting repeated concerns about its long-term viability and whether it can meet the terms of deals signed with Oracle, for example. </p><p>With this in mind, backing from Amazon – and crucially, investment – would mark a major seal of approval for the company. From Amazon’s perspective, the situation appears rather convoluted, according to Forrester principal analyst Rowan Curran. </p><p>For Amazon to justify this investment, the deal will likely have to center around heavy use of AWS infrastructure for training and inferencing purposes, as well as use of Amazon’s in-house chips. </p><p>“Amazon’s reported coming investment in OpenAI would be a tremendous expenditure of capital that will only benefit Amazon if that money goes right into OpenAI using AWS servers for training and inferencing,” he said. </p><p>From an AI model perspective, the deal doesn’t seem to add up for the hyperscaler, which already hosts OpenAI models through its Bedrock service.</p><p>Enterprises are increasingly using a mixture of options rather than sticking with a monolithic approach to the technology – a trend that AWS sought to capitalize on with the launch of Bedrock in 2023. </p><p>“The foundation model market is not commoditized, but investments of this scale into a single company do not reflect the realities of an increasingly diverse model market, and the demands of mature customers who are using an array of models to achieve their goals, rather than a single model,” Curran commented. </p><p>Similarly, the company has invested heavily in its own in-house Nova AI model range. </p><p>Amazon unveiled the launch of the Nova models at the 2025 AWS re:Invent conference in Las Vegas. As <a href="https://www.itpro.com/cloud/live/aws-re-invent-2025-all-the-news-updates-and-announcements-live-from-las-vegas"><em>ITPro </em>reported in December</a>, these models have since received an array of updates and improvements.</p><p>Notably, however, Curran suggested this could signal a “lack of confidence” in the Nova models and perhaps a growing sense of frustration at progress on this front. </p><p>“This external investment would further signal a lack of confidence in AWS's internal family of Nova models to achieve a level of functionality and usage with customers that will produce significant ROI,” he noted. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ OpenAI sets February retirement date for popular GPT-4o model ]]></title>
                                                                                                <dc:content><![CDATA[ <p>OpenAI has confirmed plans to retire one of its most popular AI models as the company reports sharp uptake of newer options. </p><p>Under plans detailed by the company this week, the much-loved <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-openais-gpt-4o-updates-including-pricing-changes-and-new-features">GPT-4o</a> model will be among those cut, as well as <a href="https://www.itpro.com/software/development/openai-wants-developers-using-its-new-gpt-4-1-models-but-how-do-they-compare-to-claude-and-gemini-on-coding-tasks">GPT-4.1</a>, GPT‑4.1 mini, and <a href="https://www.itpro.com/technology/artificial-intelligence/openai-open-weight-ai-models">OpenAI o4-mini</a>.</p><p>GPT-5 (Instant and Thinking) will also be removed, although the company previously <a href="https://openai.com/index/gpt-5-1/" target="_blank"><u>touted plans</u></a> for deprecation for both these options. </p><p>The move by OpenAI means these models will no longer be available for use in ChatGPT and comes amid a sharpened focus on pushing newer model ranges to users. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence/openai-thought-it-hit-a-home-run-with-gpt-5-users-werent-so-keen">GPT-5</a>, for example, launched in August 2025, and OpenAI has since issued a series of updates to the newer model range to keep pace with competitor options. </p><p>As <em>ITPro </em>previously reported, <a href="https://www.itpro.com/technology/artificial-intelligence/2025-marked-the-beginning-of-the-end-for-openai">CEO Sam Altman issued a ‘code red’ alert</a> at the company to push through GPT-5.2 after Google shocked the market with the launch of its more powerful <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform">Gemini 3</a> range in November. </p><h2 id="why-openai-is-retiring-gpt-4o">Why OpenAI is retiring GPT-4o</h2><p>Cutting <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-gpt-4o-including-pricing-features-and-how-to-get-access">GPT-4o</a> will come as a serious loss to some. The model was well-liked, boasting a conversational tone that proved a hit with users. </p><p>OpenAI has tried to cut GPT-4o access previously. When GPT-5 launched, access to the older model was removed, prompting a fierce backlash from users and subsequent u-turn by the company. </p><p>At the time, Altman took to X to calm users, confirming that any future plans to deprecate the model would be done so with “plenty of notice”. </p><p>In a <a href="https://help.openai.com/en/articles/20001051-retiring-gpt-4o-and-other-chatgpt-models" target="_blank"><u>blog post</u></a> detailing the plans this week, the company acknowledged there could be some frustration with removing GPT-4o. </p><p>“We know that losing access to GPT‑4o will feel frustrating for some users, and we didn’t make this decision lightly,” OpenAI said. “Retiring models is never easy, but it allows us to focus on improving the models most people use today.”</p><p>Figures published by the company also point toward waning popularity, with just 0.1% of users now opting for the older model. The “vast majority” of ChatGPT users are now sticking with GPT-5.2. </p><p>Notably, OpenAI revealed that the GPT-5-and-onward range has been influenced by learnings from GPT-4o, particularly with regard to tone and conversational style. </p><p>“We brought GPT‑4o back after hearing clear feedback from a subset of Plus and Pro users, who told us they needed more time to transition key use cases, like creative ideation, and that they preferred GPT‑4o’s conversational style and warmth,” the company said in a separate <a href="https://openai.com/index/retiring-gpt-4o-and-older-models/" target="_blank"><u>blog post</u></a>. </p><p>“That feedback directly shaped GPT‑5.1 and GPT‑5.2, with improvements to personality, stronger support for creative ideation, and more ways to customize how ChatGPT responds⁠.” </p><h2 id="when-will-access-to-gpt-4o-end">When will access to GPT-4o end?</h2><p>OpenAI revealed that all of the aforementioned models will be officially deprecated on 13 February, although access to these models via API will remain. </p><p>The company added that it will provide “advance notice” ahead of any future API retirement plans. </p><p>What this means is that conversations in ChatGPT will now automatically default to GPT-5.2 following the cut-off date. </p><p>Existing projects and conversations with older models will “remain as they are” until the cut-off date. From there one out, users will be able to continue chats using alternative models. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/openai-admits-losing-access-to-gpt-4o-will-feel-frustrating-for-users-the-company-is-pushing-ahead-with-retirement-plans-anway</link>
                                                                            <description>
                            <![CDATA[ OpenAI has confirmed plans to retire its popular GPT-4o model in February, citing increased uptake of its newer GPT-5 model range. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">NnENytbD6UXkdHjGh8MbwX</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/4CfCNLi89rRyAkt4Ut9qtV-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 30 Jan 2026 11:43:09 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:55 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/4CfCNLi89rRyAkt4Ut9qtV-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Close-up photograph of the OpenAI logo and branding on a smartphone screen, with faint warm orange lighting in background.]]></media:description>                                                            <media:text><![CDATA[Close-up photograph of the OpenAI logo and branding on a smartphone screen, with faint warm orange lighting in background.]]></media:text>
                                <media:title type="plain"><![CDATA[Close-up photograph of the OpenAI logo and branding on a smartphone screen, with faint warm orange lighting in background.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/4CfCNLi89rRyAkt4Ut9qtV-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>OpenAI has confirmed plans to retire one of its most popular AI models as the company reports sharp uptake of newer options. </p><p>Under plans detailed by the company this week, the much-loved <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-openais-gpt-4o-updates-including-pricing-changes-and-new-features">GPT-4o</a> model will be among those cut, as well as <a href="https://www.itpro.com/software/development/openai-wants-developers-using-its-new-gpt-4-1-models-but-how-do-they-compare-to-claude-and-gemini-on-coding-tasks">GPT-4.1</a>, GPT‑4.1 mini, and <a href="https://www.itpro.com/technology/artificial-intelligence/openai-open-weight-ai-models">OpenAI o4-mini</a>.</p><p>GPT-5 (Instant and Thinking) will also be removed, although the company previously <a href="https://openai.com/index/gpt-5-1/" target="_blank"><u>touted plans</u></a> for deprecation for both these options. </p><p>The move by OpenAI means these models will no longer be available for use in ChatGPT and comes amid a sharpened focus on pushing newer model ranges to users. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence/openai-thought-it-hit-a-home-run-with-gpt-5-users-werent-so-keen">GPT-5</a>, for example, launched in August 2025, and OpenAI has since issued a series of updates to the newer model range to keep pace with competitor options. </p><p>As <em>ITPro </em>previously reported, <a href="https://www.itpro.com/technology/artificial-intelligence/2025-marked-the-beginning-of-the-end-for-openai">CEO Sam Altman issued a ‘code red’ alert</a> at the company to push through GPT-5.2 after Google shocked the market with the launch of its more powerful <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform">Gemini 3</a> range in November. </p><h2 id="why-openai-is-retiring-gpt-4o">Why OpenAI is retiring GPT-4o</h2><p>Cutting <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-gpt-4o-including-pricing-features-and-how-to-get-access">GPT-4o</a> will come as a serious loss to some. The model was well-liked, boasting a conversational tone that proved a hit with users. </p><p>OpenAI has tried to cut GPT-4o access previously. When GPT-5 launched, access to the older model was removed, prompting a fierce backlash from users and subsequent u-turn by the company. </p><p>At the time, Altman took to X to calm users, confirming that any future plans to deprecate the model would be done so with “plenty of notice”. </p><p>In a <a href="https://help.openai.com/en/articles/20001051-retiring-gpt-4o-and-other-chatgpt-models" target="_blank"><u>blog post</u></a> detailing the plans this week, the company acknowledged there could be some frustration with removing GPT-4o. </p><p>“We know that losing access to GPT‑4o will feel frustrating for some users, and we didn’t make this decision lightly,” OpenAI said. “Retiring models is never easy, but it allows us to focus on improving the models most people use today.”</p><p>Figures published by the company also point toward waning popularity, with just 0.1% of users now opting for the older model. The “vast majority” of ChatGPT users are now sticking with GPT-5.2. </p><p>Notably, OpenAI revealed that the GPT-5-and-onward range has been influenced by learnings from GPT-4o, particularly with regard to tone and conversational style. </p><p>“We brought GPT‑4o back after hearing clear feedback from a subset of Plus and Pro users, who told us they needed more time to transition key use cases, like creative ideation, and that they preferred GPT‑4o’s conversational style and warmth,” the company said in a separate <a href="https://openai.com/index/retiring-gpt-4o-and-older-models/" target="_blank"><u>blog post</u></a>. </p><p>“That feedback directly shaped GPT‑5.1 and GPT‑5.2, with improvements to personality, stronger support for creative ideation, and more ways to customize how ChatGPT responds⁠.” </p><h2 id="when-will-access-to-gpt-4o-end">When will access to GPT-4o end?</h2><p>OpenAI revealed that all of the aforementioned models will be officially deprecated on 13 February, although access to these models via API will remain. </p><p>The company added that it will provide “advance notice” ahead of any future API retirement plans. </p><p>What this means is that conversations in ChatGPT will now automatically default to GPT-5.2 following the cut-off date. </p><p>Existing projects and conversations with older models will “remain as they are” until the cut-off date. From there one out, users will be able to continue chats using alternative models. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Jensen Huang says the traditional computing stack will never look the same because of AI ]]></title>
                                                                                                <dc:content><![CDATA[ <p><a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">Generative AI</a> is transforming the traditional computing stack, according to <a href="https://www.itpro.com/business/leadership/jensen-huang-has-taken-nvidia-to-the-top-but-he-says-the-suffering-wasnt-really-worth-it">Nvidia CEO Jensen Huang</a>, who says it will eventually form the core in a new era of software development.</p><p>Speaking with BlackRock CEO Larry Fink at the World Economic Forum (WEF) in Davos, Huang compared the impact of AI on modern computing to the advent of the web, the personal computing boom, and <a href="https://www.itpro.com/627952/what-is-cloud-computing">cloud computing</a>. </p><p>“It’s helpful to reason back to the first principles of fundamentally what is happening to the computing stack. This is a platform shift,” he said. </p><p>“New applications were developed to run on a new type of computer. [With the] platform shift to the internet, a new type of computing platform hosted all kinds of new applications.”</p><p>With each of these transformations, Huang noted, the “computing stack was reinvented and new applications were created”. </p><p>Generative AI essentially marks the latest step on this evolutionary path, with Huang noting that <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369965/what-is-chatgpt-and-what-does-it-mean-for-businesses">ChatGPT </a>and <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner">Anthropic’s Claude chatbot</a> will be the tip of the spear that delivers new applications and tools in the future. </p><p>“New applications will be built on top of ChatGPT,” he added. “New applications will be built on top of Anthropic Claude, for example. So it’s a platform shift in that way.”</p><h2 id="ai-is-transforming-software">AI is transforming software</h2><p>Later in his conversation with Fink, Huang noted that AI is “really easy to understand” when looking at the technology through this prism. </p><p><a href="https://www.itpro.com/strategy/28181/what-is-ai">AI</a> is helping redefine not only what constitutes modern software, but how it’s built, as enterprises flock to the technology to transform processes and streamline operations. </p><p><a href="https://www.bain.com/insights/will-agentic-ai-disrupt-saas-technology-report-2025/" target="_blank"><u>Analysis from Bain &amp; Company</u></a> shows AI is already transforming areas such as software as a service, with the arrival of agentic AI further bolstering efficiency improvements and speeding up development. </p><p>The key difference in this case is that while software in the past was “effectively pre-recorded”, AI tools rely on prompts, marking a step change in how users interact with them. They’re more intuitive, proactive, and are capable of discerning context with minimal information. </p><p>From a user efficiency perspective, this is a marked improvement when cumbersome software often relied on specific structured data to function. </p><p>“Software in the past was effectively pre-recorded. Humans would type and describe the algorithm or the recipe for the computer to execute,” he explained. “It was able to process structured information, meaning you’ve got to put the name, the address, their account number, their age, where they live.”</p><p>“You create these structured tables that software would then go and retrieve information from,” Huang added. “We call it SQL queries. SQL is the single most important database engine the world's ever known. Almost everything ran on SQL before.”</p><p>Because AI can understand unstructured information, allowing it to digest text, images, or sound, Huang said AI represents a far more convenient mode of human-computer interaction. </p><p>“For the first time, we now have a computer that is not pre-recorded, but it's processed in real-time,” he commented. “Describe it however you like to describe it, and to the extent that it can understand your intention, it could perform a task for you.”</p><h2 id="a-five-layer-cake">A “five-layer cake”</h2><p>Huang noted that the foundations upon which AI rests are crucial. It’s a delicate balancing act in which multiple “layers” need to function in tandem for the technology to operate efficiently. </p><p>This aspect adds a degree of complexity that previous technological shifts were typically exempt from. Simply put, it’s not just a simple click-and-go technology operating independently. The infrastructure stack that underpins AI is vital.</p><p>“When you think about AI, you think about the AI models, but it’s really important to understand industrially, AI is essentially a five-layer cake - at the bottom is energy,” he said.</p><p>“AI, because it's processed in real time and it generates intelligence in real time, it needs energy to do so. Energy is the first layer. The second layer is the layer that I live in. It's chips, chips and computing infrastructure.</p><p>“The next layer above it is the cloud infrastructure, the cloud services. The layer above that is the AI models.”</p><p>According to Huang, there’s a common misconception that the nitty gritty of the technology happens in the model domain. However, while this is crucial, it’s in the application layer <em>above </em>that where real value and economic growth are generated.</p><p>There are challenges here though, Huang admitted. Model performance improvements have outpaced application development, meaning organizations are struggling to unlock value from the technology. </p><p>“The reason why last year was an incredible year for AI is that the models made so much progress than the layer above it, which is ultimately the layer that we all need to succeed.”</p><p>“This application layer could be in financial services,” Huang added. “It could be in healthcare. It could be in manufacturing. This later on top, ultimately, is where economic benefit will happen.”</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/software/jensen-huang-ai-software-stack-chatgpt-claude</link>
                                                                            <description>
                            <![CDATA[ The Nvidia chief says new applications will be built “on top of ChatGPT” as the technology redefines software ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">f5iFqkcB3ipKepoqW4hgoV</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/bwwsT5qAdUjzorgSFTejTh-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 21 Jan 2026 12:15:39 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:46 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Software]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                    <category><![CDATA[Workplace Technology]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/bwwsT5qAdUjzorgSFTejTh-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Nvidia CEO Jensen Huang pictured in conversation with BlackRock CEO Larry Fink at the 2026 World Economic Forum (WEF) in Davos while wearing a black leather jacket.]]></media:description>                                                            <media:text><![CDATA[Nvidia CEO Jensen Huang pictured in conversation with BlackRock CEO Larry Fink at the 2026 World Economic Forum (WEF) in Davos while wearing a black leather jacket.]]></media:text>
                                <media:title type="plain"><![CDATA[Nvidia CEO Jensen Huang pictured in conversation with BlackRock CEO Larry Fink at the 2026 World Economic Forum (WEF) in Davos while wearing a black leather jacket.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/bwwsT5qAdUjzorgSFTejTh-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p><a href="https://www.itpro.com/technology/artificial-intelligence-ai/369959/what-is-generative-ai">Generative AI</a> is transforming the traditional computing stack, according to <a href="https://www.itpro.com/business/leadership/jensen-huang-has-taken-nvidia-to-the-top-but-he-says-the-suffering-wasnt-really-worth-it">Nvidia CEO Jensen Huang</a>, who says it will eventually form the core in a new era of software development.</p><p>Speaking with BlackRock CEO Larry Fink at the World Economic Forum (WEF) in Davos, Huang compared the impact of AI on modern computing to the advent of the web, the personal computing boom, and <a href="https://www.itpro.com/627952/what-is-cloud-computing">cloud computing</a>. </p><p>“It’s helpful to reason back to the first principles of fundamentally what is happening to the computing stack. This is a platform shift,” he said. </p><p>“New applications were developed to run on a new type of computer. [With the] platform shift to the internet, a new type of computing platform hosted all kinds of new applications.”</p><p>With each of these transformations, Huang noted, the “computing stack was reinvented and new applications were created”. </p><p>Generative AI essentially marks the latest step on this evolutionary path, with Huang noting that <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369965/what-is-chatgpt-and-what-does-it-mean-for-businesses">ChatGPT </a>and <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner">Anthropic’s Claude chatbot</a> will be the tip of the spear that delivers new applications and tools in the future. </p><p>“New applications will be built on top of ChatGPT,” he added. “New applications will be built on top of Anthropic Claude, for example. So it’s a platform shift in that way.”</p><h2 id="ai-is-transforming-software">AI is transforming software</h2><p>Later in his conversation with Fink, Huang noted that AI is “really easy to understand” when looking at the technology through this prism. </p><p><a href="https://www.itpro.com/strategy/28181/what-is-ai">AI</a> is helping redefine not only what constitutes modern software, but how it’s built, as enterprises flock to the technology to transform processes and streamline operations. </p><p><a href="https://www.bain.com/insights/will-agentic-ai-disrupt-saas-technology-report-2025/" target="_blank"><u>Analysis from Bain &amp; Company</u></a> shows AI is already transforming areas such as software as a service, with the arrival of agentic AI further bolstering efficiency improvements and speeding up development. </p><p>The key difference in this case is that while software in the past was “effectively pre-recorded”, AI tools rely on prompts, marking a step change in how users interact with them. They’re more intuitive, proactive, and are capable of discerning context with minimal information. </p><p>From a user efficiency perspective, this is a marked improvement when cumbersome software often relied on specific structured data to function. </p><p>“Software in the past was effectively pre-recorded. Humans would type and describe the algorithm or the recipe for the computer to execute,” he explained. “It was able to process structured information, meaning you’ve got to put the name, the address, their account number, their age, where they live.”</p><p>“You create these structured tables that software would then go and retrieve information from,” Huang added. “We call it SQL queries. SQL is the single most important database engine the world's ever known. Almost everything ran on SQL before.”</p><p>Because AI can understand unstructured information, allowing it to digest text, images, or sound, Huang said AI represents a far more convenient mode of human-computer interaction. </p><p>“For the first time, we now have a computer that is not pre-recorded, but it's processed in real-time,” he commented. “Describe it however you like to describe it, and to the extent that it can understand your intention, it could perform a task for you.”</p><h2 id="a-five-layer-cake">A “five-layer cake”</h2><p>Huang noted that the foundations upon which AI rests are crucial. It’s a delicate balancing act in which multiple “layers” need to function in tandem for the technology to operate efficiently. </p><p>This aspect adds a degree of complexity that previous technological shifts were typically exempt from. Simply put, it’s not just a simple click-and-go technology operating independently. The infrastructure stack that underpins AI is vital.</p><p>“When you think about AI, you think about the AI models, but it’s really important to understand industrially, AI is essentially a five-layer cake - at the bottom is energy,” he said.</p><p>“AI, because it's processed in real time and it generates intelligence in real time, it needs energy to do so. Energy is the first layer. The second layer is the layer that I live in. It's chips, chips and computing infrastructure.</p><p>“The next layer above it is the cloud infrastructure, the cloud services. The layer above that is the AI models.”</p><p>According to Huang, there’s a common misconception that the nitty gritty of the technology happens in the model domain. However, while this is crucial, it’s in the application layer <em>above </em>that where real value and economic growth are generated.</p><p>There are challenges here though, Huang admitted. Model performance improvements have outpaced application development, meaning organizations are struggling to unlock value from the technology. </p><p>“The reason why last year was an incredible year for AI is that the models made so much progress than the layer above it, which is ultimately the layer that we all need to succeed.”</p><p>“This application layer could be in financial services,” Huang added. “It could be in healthcare. It could be in manufacturing. This later on top, ultimately, is where economic benefit will happen.”</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ DeepSeek rocked Silicon Valley in January 2025 – one year on and it looks set to shake things up again ]]></title>
                                                                                                <dc:content><![CDATA[ <p>When DeepSeek hit the scene in January 2025, the Chinese AI company sent shockwaves through the industry. </p><p>The <a href="https://www.itpro.com/technology/artificial-intelligence/chinese-ai-firm-deepseek-has-silicon-valley-flustered">launch of the company’s R1 model</a>, a fine-tuned version of its previous V3 range, prompted a market meltdown. Broadcom and Nvidia shares nosedived, wiping $593 billion from the latter’s market value alone. </p><p>DeepSeek was a big wake-up call for US counterparts for a number of reasons. Chief among them was the fact it proved Chinese AI developers weren’t as far behind as some industry stakeholders believed – and the fact its advances were open source. </p><p>Notably, the most jarring aspect of the DeepSeek R1 launch lay in the reputed cost of building it. Built on a shoestring budget compared to those offered by US counterparts, this raised serious questions about the huge financial outlay associated with AI training. </p><p>In the year prior, big tech talking heads repeatedly touted the prospect of unfathomably large model training costs. Speaking in July 2024, Anthropic CEO Dario Amodei predicted skyrocketing fees in this domain, suggesting that future models <a href="https://www.itpro.com/technology/artificial-intelligence/dollar100-billion-to-build-an-ai-model-anthropic-ceo-dario-amodei-predicts-soaring-ai-training-costs-but-models-will-become-far-more-powerful"><u>could cost upwards of $100 million to build</u></a>. </p><p>It’s safe to say that <a href="https://www.itpro.com/technology/artificial-intelligence/deepseeks-r1-model-training-costs-pour-cold-water-on-big-techs-massive-ai-spending">DeepSeek poured cold water on these bold claims</a> and, in doing it with such bravado, created a sense of unease among US rivals. R1 went toe-to-toe with anything US providers could offer in terms of performance. </p><p>The model was trained on 37 billion active parameters and 671 billion total parameters, according to figures released at the time. </p><p>For context, benchmarks published by the Chinese firm showed it outpaced Anthropic’s Claude 3.5 and OpenAI’s GPT-4o in key areas, particularly in coding capabilities. </p><p>DeepSeek was an instant hit, with users globally flocking to the new application. Free to access, the DeepSeek app toppled ChatGPT as the top free application on the Apple App Store. </p><p>But one year on from the launch, where does DeepSeek stand? The initial furor and excitement has subsided to some extent, but the company is still plugging away – and so are users. </p><p><a href="https://openrouter.ai/wrapped/2025/stats" target="_blank"><u>Figures from OpenRouter</u></a> show token consumption for DeepSeek V3 0324 topped 7.27 trillion across 2025, ranking it fifth behind counterparts such as Claude Sonnet 4 and Gemini 2.0 Flash. </p><h2 id="deepseek-has-been-quietly-innovating">DeepSeek has been quietly innovating</h2><p>DeepSeek’s V3 foundation model has had a number of updates across the last year, with <a href="https://api-docs.deepseek.com/news/news250821" target="_blank"><u>V3.1 arriving in August</u></a>. A hybrid open weight model, this marked a first major step toward agentic reasoning, combining thinking and non-thinking modes within a single system. </p><p>Once again, performance and cost were key talking points, performing well on coding benchmarks while offering more efficient compute costs compared to other models on the market. </p><p>Hot on the heels of V3.1, the Chinese firm launched V3.2 and V3.2-Speciale in December. The first of these, according to DeepSeek, places a strong focus on mimicking human reasoning while the Speciale range is designed primarily for mathematical contexts, boasting "maxed-out reasoning capabilities". </p><p>Both models offered equivalent performance to <a href="https://www.itpro.com/technology/artificial-intelligence/openai-thought-it-hit-a-home-run-with-gpt-5-users-werent-so-keen">OpenAI’s GPT-5</a>, with the team describing it as “your daily driver” in a clear nod to its use in a similar fashion to <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369965/what-is-chatgpt-and-what-does-it-mean-for-businesses">ChatGPT</a>. </p><h2 id="let-s-talk-about-cybersecurity">Let’s talk about cybersecurity</h2><p>Wherever a conversation about DeepSeek arises, questions over security and data privacy haven’t been far away. Indeed, the R1 launch prompted an array of warnings from industry stakeholders about the risks associated with the model. </p><p>Indeed, February 2025 saw <a href="https://www.bbc.co.uk/news/articles/c8d95v0nr1yo" target="_blank"><u>Australian lawmakers ban the application</u></a> from government devices and systems, citing security concerns. Since then, lingering worries about the Chinese AI provider have persisted across both government and industry. </p><p>Speaking to <em>ITPro </em>in August last year, Andy Ward, SVP International at Absolute Security, said enterprises <a href="https://www.itpro.com/security/using-deepseek-at-work-security-risks"><u>should exercise extreme caution</u></a> when using the application, describing it as akin to “printing out and handing over confidential information”.</p><p>Research on the potential risks associated with DeepSeek have been highlighted in a slew of studies since its release. Less than a month after launch, researchers at <a href="https://www.itpro.com/technology/artificial-intelligence/deepseek-r1-model-jailbreak-security-flaws"><u>Cisco warned the model contained “critical safety flaws”</u></a> that left it open to an array of jailbreak techniques that could be pounced on by threat actors or leave enterprises vulnerable to data exposure. </p><p>Separate <a href="https://www.harmonic.security/resources/a-year-after-deepseek-launch-1-in-12-employees-used-china-based-ai-tools-in-the-last-month-but-kimi-moonshot-way-ahead-of-all-other-chinese-ai-tools-hiding-in-your-enterprise" target="_blank">research from Harmonic Security</a> found DeepSeek displays a disproportionately high risk of sensitive data exposure compared to other models out there on the market - even Chinese offerings. </p><p>Analysis by the company found the use of Chinese AI models is growing, with DeepSeek accounting for 25% of overall usage alongside offerings such as Alibaba’s Qwen range. </p><p>Crucially, however, DeepSeek accounts for 55% of sensitive exposure, which Harmonic researchers said was partly due to its popularity with coders. </p><h2 id="deepseek-v4-could-shake-things-up-again">DeepSeek V4 could shake things up again</h2><p>The arrival of DeepSeek V4 appears to be imminent, according to recent speculation. </p><p>Reports from <a href="https://www.theinformation.com/articles/deepseek-release-next-flagship-ai-model-strong-coding-ability" target="_blank"><u><em>The Information</em></u></a>, citing sources with intimate knowledge of the development process, said this latest iteration could arrive in early 2026. </p><p>This model will place a strong focus on code-generation features, a recurring theme in recent months amidst the arms race between Anthropic, OpenAI, and Google. </p><p>According to the publication, early benchmarks suggest the model will outperform the Claude and GPT ranges on this front. </p><p>As with the launch of DeepSeek R1, the Chinese firm is also expected to confirm another technical breakthrough on model development in terms of contending with lengthy code prompts. </p><p>If this speculation comes to fruition, we could see another scramble among US AI developers to one-up the company. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/deepseek-r1-one-year-anniversary-what-next</link>
                                                                            <description>
                            <![CDATA[ The Chinese AI company sent Silicon Valley into meltdown last year and it could rock the boat again with an upcoming model ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">PfPKtaJasGen8iPizanTTY</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/6j5SuvBk7kvPJS89t3ord5-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 20 Jan 2026 11:21:10 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:50 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/6j5SuvBk7kvPJS89t3ord5-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Logo of Chinese AI developer DeepSeek pictured on a smartphone placed on top of a laptop keyboard.]]></media:description>                                                            <media:text><![CDATA[Logo of Chinese AI developer DeepSeek pictured on a smartphone placed on top of a laptop keyboard.]]></media:text>
                                <media:title type="plain"><![CDATA[Logo of Chinese AI developer DeepSeek pictured on a smartphone placed on top of a laptop keyboard.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/6j5SuvBk7kvPJS89t3ord5-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>When DeepSeek hit the scene in January 2025, the Chinese AI company sent shockwaves through the industry. </p><p>The <a href="https://www.itpro.com/technology/artificial-intelligence/chinese-ai-firm-deepseek-has-silicon-valley-flustered">launch of the company’s R1 model</a>, a fine-tuned version of its previous V3 range, prompted a market meltdown. Broadcom and Nvidia shares nosedived, wiping $593 billion from the latter’s market value alone. </p><p>DeepSeek was a big wake-up call for US counterparts for a number of reasons. Chief among them was the fact it proved Chinese AI developers weren’t as far behind as some industry stakeholders believed – and the fact its advances were open source. </p><p>Notably, the most jarring aspect of the DeepSeek R1 launch lay in the reputed cost of building it. Built on a shoestring budget compared to those offered by US counterparts, this raised serious questions about the huge financial outlay associated with AI training. </p><p>In the year prior, big tech talking heads repeatedly touted the prospect of unfathomably large model training costs. Speaking in July 2024, Anthropic CEO Dario Amodei predicted skyrocketing fees in this domain, suggesting that future models <a href="https://www.itpro.com/technology/artificial-intelligence/dollar100-billion-to-build-an-ai-model-anthropic-ceo-dario-amodei-predicts-soaring-ai-training-costs-but-models-will-become-far-more-powerful"><u>could cost upwards of $100 million to build</u></a>. </p><p>It’s safe to say that <a href="https://www.itpro.com/technology/artificial-intelligence/deepseeks-r1-model-training-costs-pour-cold-water-on-big-techs-massive-ai-spending">DeepSeek poured cold water on these bold claims</a> and, in doing it with such bravado, created a sense of unease among US rivals. R1 went toe-to-toe with anything US providers could offer in terms of performance. </p><p>The model was trained on 37 billion active parameters and 671 billion total parameters, according to figures released at the time. </p><p>For context, benchmarks published by the Chinese firm showed it outpaced Anthropic’s Claude 3.5 and OpenAI’s GPT-4o in key areas, particularly in coding capabilities. </p><p>DeepSeek was an instant hit, with users globally flocking to the new application. Free to access, the DeepSeek app toppled ChatGPT as the top free application on the Apple App Store. </p><p>But one year on from the launch, where does DeepSeek stand? The initial furor and excitement has subsided to some extent, but the company is still plugging away – and so are users. </p><p><a href="https://openrouter.ai/wrapped/2025/stats" target="_blank"><u>Figures from OpenRouter</u></a> show token consumption for DeepSeek V3 0324 topped 7.27 trillion across 2025, ranking it fifth behind counterparts such as Claude Sonnet 4 and Gemini 2.0 Flash. </p><h2 id="deepseek-has-been-quietly-innovating">DeepSeek has been quietly innovating</h2><p>DeepSeek’s V3 foundation model has had a number of updates across the last year, with <a href="https://api-docs.deepseek.com/news/news250821" target="_blank"><u>V3.1 arriving in August</u></a>. A hybrid open weight model, this marked a first major step toward agentic reasoning, combining thinking and non-thinking modes within a single system. </p><p>Once again, performance and cost were key talking points, performing well on coding benchmarks while offering more efficient compute costs compared to other models on the market. </p><p>Hot on the heels of V3.1, the Chinese firm launched V3.2 and V3.2-Speciale in December. The first of these, according to DeepSeek, places a strong focus on mimicking human reasoning while the Speciale range is designed primarily for mathematical contexts, boasting "maxed-out reasoning capabilities". </p><p>Both models offered equivalent performance to <a href="https://www.itpro.com/technology/artificial-intelligence/openai-thought-it-hit-a-home-run-with-gpt-5-users-werent-so-keen">OpenAI’s GPT-5</a>, with the team describing it as “your daily driver” in a clear nod to its use in a similar fashion to <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369965/what-is-chatgpt-and-what-does-it-mean-for-businesses">ChatGPT</a>. </p><h2 id="let-s-talk-about-cybersecurity">Let’s talk about cybersecurity</h2><p>Wherever a conversation about DeepSeek arises, questions over security and data privacy haven’t been far away. Indeed, the R1 launch prompted an array of warnings from industry stakeholders about the risks associated with the model. </p><p>Indeed, February 2025 saw <a href="https://www.bbc.co.uk/news/articles/c8d95v0nr1yo" target="_blank"><u>Australian lawmakers ban the application</u></a> from government devices and systems, citing security concerns. Since then, lingering worries about the Chinese AI provider have persisted across both government and industry. </p><p>Speaking to <em>ITPro </em>in August last year, Andy Ward, SVP International at Absolute Security, said enterprises <a href="https://www.itpro.com/security/using-deepseek-at-work-security-risks"><u>should exercise extreme caution</u></a> when using the application, describing it as akin to “printing out and handing over confidential information”.</p><p>Research on the potential risks associated with DeepSeek have been highlighted in a slew of studies since its release. Less than a month after launch, researchers at <a href="https://www.itpro.com/technology/artificial-intelligence/deepseek-r1-model-jailbreak-security-flaws"><u>Cisco warned the model contained “critical safety flaws”</u></a> that left it open to an array of jailbreak techniques that could be pounced on by threat actors or leave enterprises vulnerable to data exposure. </p><p>Separate <a href="https://www.harmonic.security/resources/a-year-after-deepseek-launch-1-in-12-employees-used-china-based-ai-tools-in-the-last-month-but-kimi-moonshot-way-ahead-of-all-other-chinese-ai-tools-hiding-in-your-enterprise" target="_blank">research from Harmonic Security</a> found DeepSeek displays a disproportionately high risk of sensitive data exposure compared to other models out there on the market - even Chinese offerings. </p><p>Analysis by the company found the use of Chinese AI models is growing, with DeepSeek accounting for 25% of overall usage alongside offerings such as Alibaba’s Qwen range. </p><p>Crucially, however, DeepSeek accounts for 55% of sensitive exposure, which Harmonic researchers said was partly due to its popularity with coders. </p><h2 id="deepseek-v4-could-shake-things-up-again">DeepSeek V4 could shake things up again</h2><p>The arrival of DeepSeek V4 appears to be imminent, according to recent speculation. </p><p>Reports from <a href="https://www.theinformation.com/articles/deepseek-release-next-flagship-ai-model-strong-coding-ability" target="_blank"><u><em>The Information</em></u></a>, citing sources with intimate knowledge of the development process, said this latest iteration could arrive in early 2026. </p><p>This model will place a strong focus on code-generation features, a recurring theme in recent months amidst the arms race between Anthropic, OpenAI, and Google. </p><p>According to the publication, early benchmarks suggest the model will outperform the Claude and GPT ranges on this front. </p><p>As with the launch of DeepSeek R1, the Chinese firm is also expected to confirm another technical breakthrough on model development in terms of contending with lengthy code prompts. </p><p>If this speculation comes to fruition, we could see another scramble among US AI developers to one-up the company. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ OpenAI CFO touts future plans in wake of revenue gains ]]></title>
                                                                                                <dc:content><![CDATA[ <p>OpenAI has confirmed speculation that its revenue topped $20 billion this year and said the company's next goal is "practical adoption". </p><p>Heavy spending on infrastructure across the AI industry has sparked <a href="https://www.itpro.com/technology/artificial-intelligence/are-we-in-the-middle-of-a-generative-ai-bubble"><u>concerns of a bubble that may burst</u></a> before any real gains are realized from the technology. The ChatGPT developer went on a spending spree in 2025, with deals worth hundreds of billions of dollars each with <a href="https://www.itpro.com/cloud/cloud-computing/oracle-is-reaping-the-rewards-of-openais-huge-compute-requirements"><u>Oracle</u></a>, <a href="https://www.itpro.com/technology/artificial-intelligence/nvidia-openai-investment-100-billion"><u>Nvidia</u></a>, and <a href="https://www.itpro.com/technology/artificial-intelligence/openai-signs-another-chip-deal-this-time-with-amd"><u>AMD</u></a>, among <a href="https://www.itpro.com/cloud/cloud-computing/openai-just-signed-a-bumper-usd38bn-cloud-contract-with-aws-is-it-finally-preparing-to-cast-aside-microsoft"><u>others</u></a>. </p><p>OpenAI has also <a href="https://www.itpro.com/business/business-strategy/openai-1-trillion-infrastructure-investment-oracle-broadcom-amd"><u>reportedly sought new ways to boost revenue streams</u></a>, announcing plans last week to start showing adverts to some ChatGPT users. </p><p>Sarah Friar, OpenAI's chief financial officer, said in a <a href="https://openai.com/index/a-business-that-scales-with-the-value-of-intelligence/" target="_blank"><u>blog post</u></a> that the company's revenue topped $20 billion in 2025, suggesting it was tripling annually to match the rate of compute available. </p><p>"Looking back on the past three years, our ability to serve customers—as measured by revenue—directly tracks available compute: Compute grew 3X year over year or 9.5X from 2023 to 2025: 0.2 GW in 2023, 0.6 GW in 2024, and ~1.9 GW in 2025," she said. </p><p>"While revenue followed the same curve growing 3X year over year, or 10X from 2023 to 2025: $2B ARR [annualized run rate] in 2023, $6B in 2024, and $20B+ in 2025."</p><p>She added: "This is never-before-seen growth at such scale. And we firmly believe that more compute in these periods would have led to faster customer adoption and monetization."</p><h2 id="openai-needs-more-compute-power">OpenAI needs more compute power</h2><p>Friar suggested that a drive for more compute capacity was one reason that OpenAI's <a href="https://www.itpro.com/technology/artificial-intelligence/the-honeymoon-period-is-officially-over-for-microsoft-and-openai">exclusive partnership with Microsoft</a> ended last year as the company sought to widen provider options.</p><p>"Three years ago, we relied on a single compute provider," she noted. "Today, we are working with providers across a diversified ecosystem. That shift gives us resilience and, critically, compute certainty. We can plan, finance, and deploy capacity with confidence in a market where access to compute defines who can scale."</p><p>She suggested the aim of scaling compute was to keep down the costs of AI workloads, to make access to the technology viable for "everyday" users. </p><p>Exactly how people use ChatGPT has been a point of contention over the last two years, particularly as the company looks to drive enterprise adoption of its AI solutions. <a href="https://www.itpro.com/technology/artificial-intelligence/openai-just-revealed-what-people-really-use-chatgpt-for-and-70-percent-of-queries-have-nothing-to-do-with-work"><u>OpenAI-backed research</u></a> suggested more than 70% of ChatGPT queries have nothing to do with work, but are about how to write or personal questions. </p><p>The next big hurdle, Friar noted, will be bolstering “practical adoption” of the technology across both the consumer and enterprise domains. </p><p>"The priority is closing the gap between what AI now makes possible and how people, companies, and countries are using it day to day," she said. "The opportunity is large and immediate, especially in health, science, and enterprise, where better intelligence translates directly into better outcomes."</p><p>From a technology standpoint, she said the next phase for AI is agents and automation that can take action, run continuously and maintain context over time. </p><p>"For individuals, that means AI that manages projects, coordinates plans, and executes tasks," Friar said. "For organizations, it becomes an operating layer for knowledge work."</p><h2 id="how-is-openai-paying-for-it-all">How is OpenAI paying for it all?</h2><p>OpenAI has made no secret of its lavish spending over the last 18 months and it faces a growing list of bills. To compensate for this, Friar added that new business models will emerge, including licensing, IP-based agreements and "outcome-based pricing". </p><p>What this suggests is that OpenAI is hoping to earn revenue from what partners create using its tools. </p><p>Beyond that, OpenAI last week said it was introducing advertising in ChatGPT for some users – notably anyone using the app for free or with a budget ChatGPT GO subscription – to further expand its revenue sources. </p><p>In a <a href="https://openai.com/index/our-approach-to-advertising-and-expanding-access/" target="_blank">blog post</a> detailing the announcement, the company said the ads will be kept separate from the chat bot and it will "never sell your data" to advertisers, who won't be able to influence answers to queries. </p><p>OpenAI's policy chief Chris Lehane told <a href="https://www.axios.com/2026/01/19/openai-device-2026-lehane-jony-ive" target="_blank"><u><em>Axios</em></u></a><em> </em>that the company is "on track" to release an AI device in the second half of this year, widely expected after the <a href="https://www.itpro.com/hardware/openai-just-made-its-first-hardware-acquisition-with-a-deal-for-jony-ives-startup-but-altmans-usd6-4-billion-moonshot-is-a-huge-risk">acquisition of Jonny Ive's hardware startup</a> io Products last year. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/business/business-strategy/openai-cfo-teases-new-money-making-plans-as-ai-giant-targets-practical-adoption-gains-in-2026-and-beyond</link>
                                                                            <description>
                            <![CDATA[ OpenAI has confirmed speculation that its revenue topped $20 billion this year and said the company's next goal is "practical adoption" of its AI solutions across both the consumer and enterprise domains. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">gK2T73zpiAqU9FDAogW3YD</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/jqwWbYxM9J7XrKoJg5Lk3S-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 20 Jan 2026 11:10:33 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 15:00:15 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/jqwWbYxM9J7XrKoJg5Lk3S-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[OpenAI chief financial officer (CFO) Sarah Friar pictured during an interview at Bloomberg House during the World Economic Forum (WEF) in Davos, Switzerland. ]]></media:description>                                                            <media:text><![CDATA[OpenAI chief financial officer (CFO) Sarah Friar pictured during an interview at Bloomberg House during the World Economic Forum (WEF) in Davos, Switzerland. ]]></media:text>
                                <media:title type="plain"><![CDATA[OpenAI chief financial officer (CFO) Sarah Friar pictured during an interview at Bloomberg House during the World Economic Forum (WEF) in Davos, Switzerland. ]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/jqwWbYxM9J7XrKoJg5Lk3S-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>OpenAI has confirmed speculation that its revenue topped $20 billion this year and said the company's next goal is "practical adoption". </p><p>Heavy spending on infrastructure across the AI industry has sparked <a href="https://www.itpro.com/technology/artificial-intelligence/are-we-in-the-middle-of-a-generative-ai-bubble"><u>concerns of a bubble that may burst</u></a> before any real gains are realized from the technology. The ChatGPT developer went on a spending spree in 2025, with deals worth hundreds of billions of dollars each with <a href="https://www.itpro.com/cloud/cloud-computing/oracle-is-reaping-the-rewards-of-openais-huge-compute-requirements"><u>Oracle</u></a>, <a href="https://www.itpro.com/technology/artificial-intelligence/nvidia-openai-investment-100-billion"><u>Nvidia</u></a>, and <a href="https://www.itpro.com/technology/artificial-intelligence/openai-signs-another-chip-deal-this-time-with-amd"><u>AMD</u></a>, among <a href="https://www.itpro.com/cloud/cloud-computing/openai-just-signed-a-bumper-usd38bn-cloud-contract-with-aws-is-it-finally-preparing-to-cast-aside-microsoft"><u>others</u></a>. </p><p>OpenAI has also <a href="https://www.itpro.com/business/business-strategy/openai-1-trillion-infrastructure-investment-oracle-broadcom-amd"><u>reportedly sought new ways to boost revenue streams</u></a>, announcing plans last week to start showing adverts to some ChatGPT users. </p><p>Sarah Friar, OpenAI's chief financial officer, said in a <a href="https://openai.com/index/a-business-that-scales-with-the-value-of-intelligence/" target="_blank"><u>blog post</u></a> that the company's revenue topped $20 billion in 2025, suggesting it was tripling annually to match the rate of compute available. </p><p>"Looking back on the past three years, our ability to serve customers—as measured by revenue—directly tracks available compute: Compute grew 3X year over year or 9.5X from 2023 to 2025: 0.2 GW in 2023, 0.6 GW in 2024, and ~1.9 GW in 2025," she said. </p><p>"While revenue followed the same curve growing 3X year over year, or 10X from 2023 to 2025: $2B ARR [annualized run rate] in 2023, $6B in 2024, and $20B+ in 2025."</p><p>She added: "This is never-before-seen growth at such scale. And we firmly believe that more compute in these periods would have led to faster customer adoption and monetization."</p><h2 id="openai-needs-more-compute-power">OpenAI needs more compute power</h2><p>Friar suggested that a drive for more compute capacity was one reason that OpenAI's <a href="https://www.itpro.com/technology/artificial-intelligence/the-honeymoon-period-is-officially-over-for-microsoft-and-openai">exclusive partnership with Microsoft</a> ended last year as the company sought to widen provider options.</p><p>"Three years ago, we relied on a single compute provider," she noted. "Today, we are working with providers across a diversified ecosystem. That shift gives us resilience and, critically, compute certainty. We can plan, finance, and deploy capacity with confidence in a market where access to compute defines who can scale."</p><p>She suggested the aim of scaling compute was to keep down the costs of AI workloads, to make access to the technology viable for "everyday" users. </p><p>Exactly how people use ChatGPT has been a point of contention over the last two years, particularly as the company looks to drive enterprise adoption of its AI solutions. <a href="https://www.itpro.com/technology/artificial-intelligence/openai-just-revealed-what-people-really-use-chatgpt-for-and-70-percent-of-queries-have-nothing-to-do-with-work"><u>OpenAI-backed research</u></a> suggested more than 70% of ChatGPT queries have nothing to do with work, but are about how to write or personal questions. </p><p>The next big hurdle, Friar noted, will be bolstering “practical adoption” of the technology across both the consumer and enterprise domains. </p><p>"The priority is closing the gap between what AI now makes possible and how people, companies, and countries are using it day to day," she said. "The opportunity is large and immediate, especially in health, science, and enterprise, where better intelligence translates directly into better outcomes."</p><p>From a technology standpoint, she said the next phase for AI is agents and automation that can take action, run continuously and maintain context over time. </p><p>"For individuals, that means AI that manages projects, coordinates plans, and executes tasks," Friar said. "For organizations, it becomes an operating layer for knowledge work."</p><h2 id="how-is-openai-paying-for-it-all">How is OpenAI paying for it all?</h2><p>OpenAI has made no secret of its lavish spending over the last 18 months and it faces a growing list of bills. To compensate for this, Friar added that new business models will emerge, including licensing, IP-based agreements and "outcome-based pricing". </p><p>What this suggests is that OpenAI is hoping to earn revenue from what partners create using its tools. </p><p>Beyond that, OpenAI last week said it was introducing advertising in ChatGPT for some users – notably anyone using the app for free or with a budget ChatGPT GO subscription – to further expand its revenue sources. </p><p>In a <a href="https://openai.com/index/our-approach-to-advertising-and-expanding-access/" target="_blank">blog post</a> detailing the announcement, the company said the ads will be kept separate from the chat bot and it will "never sell your data" to advertisers, who won't be able to influence answers to queries. </p><p>OpenAI's policy chief Chris Lehane told <a href="https://www.axios.com/2026/01/19/openai-device-2026-lehane-jony-ive" target="_blank"><u><em>Axios</em></u></a><em> </em>that the company is "on track" to release an AI device in the second half of this year, widely expected after the <a href="https://www.itpro.com/hardware/openai-just-made-its-first-hardware-acquisition-with-a-deal-for-jony-ives-startup-but-altmans-usd6-4-billion-moonshot-is-a-huge-risk">acquisition of Jonny Ive's hardware startup</a> io Products last year. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Are AI browsers a golden opportunity or  cybersecurity nightmare? ]]></title>
                                                                                                <dc:content><![CDATA[ <p>It feels as if nothing in the world of technology is truly free from the influence of AI, generative AI especially. </p><p>Last year gave rise to the first cohort of <a href="https://www.itpro.com/technology/artificial-intelligence/a-threat-to-googles-dominance-the-ai-browser-wars-have-begun-here-are-the-top-contenders-vying-for-the-crown"><u>AI browsers</u></a> – a new battleground for AI dominance. Simply put, AI is integrated into web browsers so users can benefit from features like article summaries, automation, and increased personalization.  </p><p>But as with any new tool, cyber criminals are rubbing their hands with glee over a new avenue by which they can terrorize users and put them at risk. Threats like prompt injection, especially, are prevalent. </p><p>Enterprises may therefore be asking whether these new tools can unlock additional productivity gains or whether the risks, many of which might not have yet been realized, outweigh these early benefits. With AI browsers having only just emerged, do they have a place in business?</p><div><blockquote><p>"The risk isn't that employees will do something stupid. It's that the AI will do something fast."</p></blockquote></div><h2 id="what-are-ai-browsers-and-how-useful-are-they">What are AI browsers and how useful are they? </h2><p>AI browsers may seem like an inevitability in the age of AI; why use separate <a href="https://www.itpro.com/technology/artificial-intelligence/practical-ai-the-age-of-agentic-ai"><u>AI agents</u></a> or digital assistants for internet-facing work when you can use a single application that integrates the best of all worlds? </p><p>Examples of strong use cases, according to cybersecurity and AI analyst at Keirstone Limited Francis Fabrizi, include not just searching for keywords but looking across all open tabs and synthesizing information. For example, AI browsers can offer time-savings in the form of instant summaries of long meetings or even "agentic" capabilities like booking travel within company policy.</p><p>"The real draw of AI browsers for professionals is that they stop being just a window to the web and start acting like a digital assistant that understands context," he tells <em>ITPro</em>. "For a business, this means getting through "drudge work" significantly faster which is why the adoption rate is so high.”</p><p>For Rob T Lee, chief of research and <a href="https://www.itpro.com/business/business-strategy/are-chief-ai-officers-here-to-stay">chief AI officer</a> at SANS Institute, the appeal is "watching the browser do your job while you drink coffee". </p><p>"These agents can [literally] take over your active tab and execute tasks," he says. "Open Wordle, tell the AI to solve it, and watch it physically click the keys while you sit there. That's the party trick."</p><p>Beyond the simple tasks, there's "telepathic stuff" like highlighting a confusing paragraph in a technical PDF or a code snippet on <a href="https://www.itpro.com/software/development/github-just-launched-a-new-mission-control-center-for-developers-to-delegate-tasks-to-ai-coding-agents">GitHub</a>, with the assistant button paving the way for instant explanations and simplifications. But does any of this come at a cost? "The productivity gains are real," he advocates. "But the question is <a href="https://www.itpro.com/security/agentic-ai-poses-major-challenge-for-security-professionals-says-palo-alto-networks-emea-ciso">whether your security architecture can handle an autonomous agent </a>with that much access."</p><h2 id="the-risks-of-ai-browsers-for-businesses">The risks of AI browsers for businesses</h2><p><a href="https://www.itpro.com/technology/artificial-intelligence/openai-chatgpt-atlas-ai-browser-prompt-injection-attack-risk"><u>Prompt injection attacks</u></a> are far and away the most prominent threat for AI browsers. This manifests when a malicious site hides invisible instructions that the AI reads while summarizing a page. They could trick your browser into exfiltrating private data, for example, or stealing session cookies. "Because AI browsers have more "autonomy" than traditional ones, an attacker doesn't necessarily need to hack your computer, they just need to trick the AI into following a bad command," says Fabrizi. </p><p>There's also OAuth abuse, adds Lee. "When users grant AI browsers access to Gmail, Calendar, or Drive, attackers exploiting prompt injection can extract emails, create forwarding rules, harvest tokens, and all without stealing a password. The agent already has the keys." </p><iframe allow="" height="200px" width="100%" id="" style="" class="position-center" data-lazy-priority="low" data-lazy-src="https://player.captivate.fm/episode/d27ed604-4541-4b22-acce-3c5ab40e5fc9/" src="https://player.captivate.fm/episode/d27ed604-4541-4b22-acce-3c5ab40e5fc9/"></iframe><p>Beyond cybersecurity, users also risk data leakage – or losing control of sensitive information when data is beamed out to external servers. That sensitive data or proprietary IP might be used to train future models. Workers may also become complacent, adds Fabrizi, relying on AI to handle routine tasks that they neglect to double check. Hallucinations may slip through the cracks and factual errors may wind up being factored into serious business decisions much later down the line.  </p><p>Web browsers already contain sensitive data like history, cookies, session data, cross-application workflows, adds Lee. "Layer an AI agent on top of that with permission to act on your behalf, and you've created something with enormous access operating at a speed and scale no human could match.</p><p>"The risk isn't that employees will do something stupid. It's that the AI will do something fast."</p><h2 id="how-can-businesses-safely-implement-ai-browsers">How can businesses safely implement AI browsers?</h2><p>There's no doubt that AI browsers are in the ascendancy – whether or not they're currently safe or dangerous, employees will be inevitably drawn to the benefits. Implementing blanket bans, therefore, <a href="https://www.itpro.com/technology/artificial-intelligence/shadow-ai-can-be-a-tool-for-ai-innovation-with-the-right-controls-say-gartner-analysts">will likely be counterintuitive</a>. Both Fabrizi and Lee agree that CISOs and IT leaders should not default to bans or blocking, as this will lead to yet more <a href="https://www.itpro.com/technology/artificial-intelligence/gartner-says-40-percent-of-enterprises-will-experience-shadow-ai-breaches-by-2030-educating-staff-is-the-key-to-avoiding-disaster">"shadow AI"</a>.</p><p>"Instead, the approach should be a managed rollout that prioritizes data loss prevention and strict <a href="https://www.itpro.com/technology/artificial-intelligence/should-workers-prepare-to-become-ai-agent-bosses">"human-in-the-loop"</a> requirements," says Fabrizi. "There is absolutely a valid use case for AI browsers in research, content creation and data analysis but it has to be done through enterprise-grade versions where the company retains ownership of the data."</p><p>Whether or not the risk outweighs the benefit is not a question of the existence of the tool itself, but rather how it's implemented within an organization and how it's ultimately used. IT leaders should focus on how to <a href="https://www.itpro.com/technology/artificial-intelligence/enterprises-are-worried-about-agentic-ai-security-risks-gartner-says-the-answer-is-just-adding-more-ai-agents">implement proper guardrails</a> while preserving employees' desire to extract more productivity gains from the new technology.</p><p>One example includes disabling the AI's ability to act autonomously (like sending emails or moving files with manual input), Fabrizi advises. Lee adds that organizations should classify data to gain an understanding of what's actually sensitive. </p><p>Identifying enterprise-grade AI browsers with contractual data protection measures will be a great starting point when assessing which service to adopt. Lee also advises implementing controls around OAuth scopes and session management, while running pilots with users who don't have access to the most sensitive corporate or organization data.</p><p>"Valid use cases absolutely exist and the productivity gains are real," Lee says. "But you need visibility first. Bring shadow AI into the sunlight, then secure it."</p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/ai-browsers-golden-opportunity-for-business-or-cybersecurity-nightmare</link>
                                                                            <description>
                            <![CDATA[ AI browsers are on the rise despite the concrete risks associated with using them ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">3y6YTPT4eRC59cTzzo7fwP</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/sztp9b6MMEKU7iZkg4UAvA-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 15 Jan 2026 12:30:56 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:16 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ keumars.afifi-sabet@futurenet.com (Keumars Afifi-Sabet) ]]></author>                    <dc:creator><![CDATA[ Keumars Afifi-Sabet ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/EAvwpZggMZ2K5h8s2pTAEm-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/sztp9b6MMEKU7iZkg4UAvA-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A stylized graphic showing the outline of a laptop with web browser, overlaid with yellow stickers depicting an AI agent, a coding window, a dial, and a switch, to represent AI browsers, on a light blue background.]]></media:description>                                                            <media:text><![CDATA[A stylized graphic showing the outline of a laptop with web browser, overlaid with yellow stickers depicting an AI agent, a coding window, a dial, and a switch, to represent AI browsers, on a light blue background.]]></media:text>
                                <media:title type="plain"><![CDATA[A stylized graphic showing the outline of a laptop with web browser, overlaid with yellow stickers depicting an AI agent, a coding window, a dial, and a switch, to represent AI browsers, on a light blue background.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/sztp9b6MMEKU7iZkg4UAvA-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>It feels as if nothing in the world of technology is truly free from the influence of AI, generative AI especially. </p><p>Last year gave rise to the first cohort of <a href="https://www.itpro.com/technology/artificial-intelligence/a-threat-to-googles-dominance-the-ai-browser-wars-have-begun-here-are-the-top-contenders-vying-for-the-crown"><u>AI browsers</u></a> – a new battleground for AI dominance. Simply put, AI is integrated into web browsers so users can benefit from features like article summaries, automation, and increased personalization.  </p><p>But as with any new tool, cyber criminals are rubbing their hands with glee over a new avenue by which they can terrorize users and put them at risk. Threats like prompt injection, especially, are prevalent. </p><p>Enterprises may therefore be asking whether these new tools can unlock additional productivity gains or whether the risks, many of which might not have yet been realized, outweigh these early benefits. With AI browsers having only just emerged, do they have a place in business?</p><div><blockquote><p>"The risk isn't that employees will do something stupid. It's that the AI will do something fast."</p></blockquote></div><h2 id="what-are-ai-browsers-and-how-useful-are-they">What are AI browsers and how useful are they? </h2><p>AI browsers may seem like an inevitability in the age of AI; why use separate <a href="https://www.itpro.com/technology/artificial-intelligence/practical-ai-the-age-of-agentic-ai"><u>AI agents</u></a> or digital assistants for internet-facing work when you can use a single application that integrates the best of all worlds? </p><p>Examples of strong use cases, according to cybersecurity and AI analyst at Keirstone Limited Francis Fabrizi, include not just searching for keywords but looking across all open tabs and synthesizing information. For example, AI browsers can offer time-savings in the form of instant summaries of long meetings or even "agentic" capabilities like booking travel within company policy.</p><p>"The real draw of AI browsers for professionals is that they stop being just a window to the web and start acting like a digital assistant that understands context," he tells <em>ITPro</em>. "For a business, this means getting through "drudge work" significantly faster which is why the adoption rate is so high.”</p><p>For Rob T Lee, chief of research and <a href="https://www.itpro.com/business/business-strategy/are-chief-ai-officers-here-to-stay">chief AI officer</a> at SANS Institute, the appeal is "watching the browser do your job while you drink coffee". </p><p>"These agents can [literally] take over your active tab and execute tasks," he says. "Open Wordle, tell the AI to solve it, and watch it physically click the keys while you sit there. That's the party trick."</p><p>Beyond the simple tasks, there's "telepathic stuff" like highlighting a confusing paragraph in a technical PDF or a code snippet on <a href="https://www.itpro.com/software/development/github-just-launched-a-new-mission-control-center-for-developers-to-delegate-tasks-to-ai-coding-agents">GitHub</a>, with the assistant button paving the way for instant explanations and simplifications. But does any of this come at a cost? "The productivity gains are real," he advocates. "But the question is <a href="https://www.itpro.com/security/agentic-ai-poses-major-challenge-for-security-professionals-says-palo-alto-networks-emea-ciso">whether your security architecture can handle an autonomous agent </a>with that much access."</p><h2 id="the-risks-of-ai-browsers-for-businesses">The risks of AI browsers for businesses</h2><p><a href="https://www.itpro.com/technology/artificial-intelligence/openai-chatgpt-atlas-ai-browser-prompt-injection-attack-risk"><u>Prompt injection attacks</u></a> are far and away the most prominent threat for AI browsers. This manifests when a malicious site hides invisible instructions that the AI reads while summarizing a page. They could trick your browser into exfiltrating private data, for example, or stealing session cookies. "Because AI browsers have more "autonomy" than traditional ones, an attacker doesn't necessarily need to hack your computer, they just need to trick the AI into following a bad command," says Fabrizi. </p><p>There's also OAuth abuse, adds Lee. "When users grant AI browsers access to Gmail, Calendar, or Drive, attackers exploiting prompt injection can extract emails, create forwarding rules, harvest tokens, and all without stealing a password. The agent already has the keys." </p><iframe allow="" height="200px" width="100%" id="" style="" class="position-center" data-lazy-priority="low" data-lazy-src="https://player.captivate.fm/episode/d27ed604-4541-4b22-acce-3c5ab40e5fc9/" src="https://player.captivate.fm/episode/d27ed604-4541-4b22-acce-3c5ab40e5fc9/"></iframe><p>Beyond cybersecurity, users also risk data leakage – or losing control of sensitive information when data is beamed out to external servers. That sensitive data or proprietary IP might be used to train future models. Workers may also become complacent, adds Fabrizi, relying on AI to handle routine tasks that they neglect to double check. Hallucinations may slip through the cracks and factual errors may wind up being factored into serious business decisions much later down the line.  </p><p>Web browsers already contain sensitive data like history, cookies, session data, cross-application workflows, adds Lee. "Layer an AI agent on top of that with permission to act on your behalf, and you've created something with enormous access operating at a speed and scale no human could match.</p><p>"The risk isn't that employees will do something stupid. It's that the AI will do something fast."</p><h2 id="how-can-businesses-safely-implement-ai-browsers">How can businesses safely implement AI browsers?</h2><p>There's no doubt that AI browsers are in the ascendancy – whether or not they're currently safe or dangerous, employees will be inevitably drawn to the benefits. Implementing blanket bans, therefore, <a href="https://www.itpro.com/technology/artificial-intelligence/shadow-ai-can-be-a-tool-for-ai-innovation-with-the-right-controls-say-gartner-analysts">will likely be counterintuitive</a>. Both Fabrizi and Lee agree that CISOs and IT leaders should not default to bans or blocking, as this will lead to yet more <a href="https://www.itpro.com/technology/artificial-intelligence/gartner-says-40-percent-of-enterprises-will-experience-shadow-ai-breaches-by-2030-educating-staff-is-the-key-to-avoiding-disaster">"shadow AI"</a>.</p><p>"Instead, the approach should be a managed rollout that prioritizes data loss prevention and strict <a href="https://www.itpro.com/technology/artificial-intelligence/should-workers-prepare-to-become-ai-agent-bosses">"human-in-the-loop"</a> requirements," says Fabrizi. "There is absolutely a valid use case for AI browsers in research, content creation and data analysis but it has to be done through enterprise-grade versions where the company retains ownership of the data."</p><p>Whether or not the risk outweighs the benefit is not a question of the existence of the tool itself, but rather how it's implemented within an organization and how it's ultimately used. IT leaders should focus on how to <a href="https://www.itpro.com/technology/artificial-intelligence/enterprises-are-worried-about-agentic-ai-security-risks-gartner-says-the-answer-is-just-adding-more-ai-agents">implement proper guardrails</a> while preserving employees' desire to extract more productivity gains from the new technology.</p><p>One example includes disabling the AI's ability to act autonomously (like sending emails or moving files with manual input), Fabrizi advises. Lee adds that organizations should classify data to gain an understanding of what's actually sensitive. </p><p>Identifying enterprise-grade AI browsers with contractual data protection measures will be a great starting point when assessing which service to adopt. Lee also advises implementing controls around OAuth scopes and session management, while running pilots with users who don't have access to the most sensitive corporate or organization data.</p><p>"Valid use cases absolutely exist and the productivity gains are real," Lee says. "But you need visibility first. Bring shadow AI into the sunlight, then secure it."</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Anthropic’s Claude AI chatbot is down as company confirms ‘elevated error rates’ for Opus 4.5 and Sonnet 4.5 ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Anthropic’s Claude AI is down, with users reporting issues accessing the popular chatbot. </p><p>Reports of an outage have been circulating social media since around 9am (UK time) today, with many encountering errors while using it and others unable to access the tool entirely. </p><p>According to <a href="https://downdetector.co.uk/status/claude-ai/" target="_blank">figures from Downdetector</a>, there have been more than 1,500 separate reports highlighting issues with the platform. </p><p>Anthropic has confirmed the issue, advising users that they may encounter “elevated error rates” on Sonnet 4.5 and Opus 4.5, two of the company’s leading AI models. </p><p>Opus 4.5 <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner"><u>was launched in November last year</u></a>, and primarily used in coding activities by software developers.</p><p>Exact details on the cause of the outage are yet to be determined. However, in a status update, Anthropic said the issue had been “identified and a fix is being implemented”. </p><p>That update, issued at 9.16am (UK time), has since been followed up by three separate updates insisting the company is “continuing to work on a fix for this issue”, suggesting continued difficulties for the firm.</p><p><em>ITPro </em>has verified the issues, having first encountered issues accessing the service before being met with a message warning responses are “taking longer than usual”.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:1920px;"><p class="vanilla-image-block" style="padding-top:48.33%;"><img id="dYPTnfvimLWf7UJbkLh4GR" name="claude down" alt="Outage advisory message on Claude AI chatbot." src="https://cdn.mos.cms.futurecdn.net/dYPTnfvimLWf7UJbkLh4GR-1920-80.jpg" mos="" align="middle" fullscreen="" width="1920" height="928" attribution="" endorsement="" class="inline" /></p></div></div><figcaption itemprop="caption description" class=" inline-layout"><span class="credit" itemprop="copyrightHolder">(Image credit: ITPro/Rory Bathgate)</span></figcaption></figure><p><strong>UPDATE</strong>: Anthropic had announced a fix has been implemented and will continue "monitoring the results". </p><p>Statistics from Downdetector now show a marked dip in outage reports.  </p><p>This is a developing story and <em>ITPro </em>will update as new information becomes available. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/anthropic-claude-outage-opus-sonnet-down</link>
                                                                            <description>
                            <![CDATA[ Users of Anthropic's Sonnet 4.5 and Opus 4.5 models are being met with "elevated error rates" ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">C7Dtjjrv6kiovZ3zk5ZioU</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/2oHsAFiP3SKDjPon8yDXqN-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 14 Jan 2026 11:53:11 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:48 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/2oHsAFiP3SKDjPon8yDXqN-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Anthropic Claude logo and branding pictured on a smartphone screen held in hand, with separate branding pictured in background.]]></media:description>                                                            <media:text><![CDATA[Anthropic Claude logo and branding pictured on a smartphone screen held in hand, with separate branding pictured in background.]]></media:text>
                                <media:title type="plain"><![CDATA[Anthropic Claude logo and branding pictured on a smartphone screen held in hand, with separate branding pictured in background.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/2oHsAFiP3SKDjPon8yDXqN-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Anthropic’s Claude AI is down, with users reporting issues accessing the popular chatbot. </p><p>Reports of an outage have been circulating social media since around 9am (UK time) today, with many encountering errors while using it and others unable to access the tool entirely. </p><p>According to <a href="https://downdetector.co.uk/status/claude-ai/" target="_blank">figures from Downdetector</a>, there have been more than 1,500 separate reports highlighting issues with the platform. </p><p>Anthropic has confirmed the issue, advising users that they may encounter “elevated error rates” on Sonnet 4.5 and Opus 4.5, two of the company’s leading AI models. </p><p>Opus 4.5 <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner"><u>was launched in November last year</u></a>, and primarily used in coding activities by software developers.</p><p>Exact details on the cause of the outage are yet to be determined. However, in a status update, Anthropic said the issue had been “identified and a fix is being implemented”. </p><p>That update, issued at 9.16am (UK time), has since been followed up by three separate updates insisting the company is “continuing to work on a fix for this issue”, suggesting continued difficulties for the firm.</p><p><em>ITPro </em>has verified the issues, having first encountered issues accessing the service before being met with a message warning responses are “taking longer than usual”.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:1920px;"><p class="vanilla-image-block" style="padding-top:48.33%;"><img id="dYPTnfvimLWf7UJbkLh4GR" name="claude down" alt="Outage advisory message on Claude AI chatbot." src="https://cdn.mos.cms.futurecdn.net/dYPTnfvimLWf7UJbkLh4GR-1920-80.jpg" mos="" align="middle" fullscreen="" width="1920" height="928" attribution="" endorsement="" class="inline" /></p></div></div><figcaption itemprop="caption description" class=" inline-layout"><span class="credit" itemprop="copyrightHolder">(Image credit: ITPro/Rory Bathgate)</span></figcaption></figure><p><strong>UPDATE</strong>: Anthropic had announced a fix has been implemented and will continue "monitoring the results". </p><p>Statistics from Downdetector now show a marked dip in outage reports.  </p><p>This is a developing story and <em>ITPro </em>will update as new information becomes available. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Google's Apple deal shows it's pulling ahead in the AI race ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Apple is turning to Google to help underpin its Siri service and “Apple Intelligence” capabilities, marking the latest sign the tech giant is gaining momentum in the AI race. </p><p>The duo have signed a multi-year partnership that will see Google support a major upgrade to Siri, which is expected to launch later this year. </p><p>Moving ahead, Apple’s foundation models will also be based on Google’s in-house Gemini AI model range, the company confirmed. </p><p>“After careful evaluation, Apple determined that Google's Al technology provides the most capable foundation for Apple Foundation Models and is excited about the innovative new experiences it will unlock for Apple users,” the duo said in a joint statement confirming the deal. </p><p>“Apple Intelligence will continue to run on Apple devices and Private Cloud Compute, while maintaining Apple's industry-leading privacy standards.”</p><p>The move comes in the wake of several months of speculation over closer ties between the two tech giants. Google already pays Apple billions of dollars to remain the default search engine on iPhones, for example. </p><p>According to reports from <a href="https://www.bloomberg.com/news/articles/2025-08-22/apple-explores-using-google-gemini-ai-to-power-revamped-siri?" target="_blank"><u><em>Bloomberg</em></u></a> in August last year, talks between the two over a potential deal for Gemini to power Siri were in the works. Separate <a href="https://www.bloomberg.com/news/articles/2025-11-05/apple-plans-to-use-1-2-trillion-parameter-google-gemini-model-to-power-new-siri" target="_blank"><u>reports </u></a>from the publication noted the deal could see Apple pay Google upwards of £1 billion a year to use the model range. </p><p>The deal also represents somewhat of a snub for OpenAI. Apple previously integrated ChatGPT into macOS and forged close ties with the company on Apple Intelligence, for example. </p><p>Opting for a deal with Google suggests the tech giant views the Gemini range as a more reliable option than OpenAI's flagship GPT series. </p><h2 id="google-is-pulling-ahead-in-the-ai-race">Google is pulling ahead in the AI race</h2><p>From Google’s perspective, the deal with Apple marks a huge seal of approval for its Gemini model range. Wedbush analyst Dan Ives said the partnership is a “major validation moment for Google” in a Monday research note, per reports from <a href="https://fortune.com/2026/01/12/apple-intelligence-powered-by-gemini-marks-a-major-validation-moment-for-google-top-tech-analyst-says/"><u><em>Fortune</em></u></a>. </p><p>Alphabet, Google’s parent company, hit a major milestone in the wake of the deal, surpassing a $4 trillion valuation for the first time. The surge means Alphabet is just the fourth company to hit such a milestone alongside Nvidia, Microsoft, and Apple. </p><p>A key factor behind the breakneck growth at Google has been its growth strength in the generative AI race. The company has recorded consistent growth and progress on this front over the last three years following a rocky start. </p><p>When <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369965/what-is-chatgpt-and-what-does-it-mean-for-businesses">ChatGPT </a>launched in November 2022, Google was caught off guard by the release, prompting a rapid shift in focus led by CEO Sundar Pichai. </p><p>As <em>ITPro </em><a href="https://www.itpro.com/technology/artificial-intelligence-ai/369766/google-upends-teams-to-counter-threat-chatgpt"><u>reported at the time</u></a>, CEO Sundar Pichai issued a ‘code red’ alert which saw the company reallocate personnel and resources in a bid to keep pace with OpenAI and Microsoft. </p><p>Recent events suggest that Google has firmly caught up with key competitors in the AI space, particularly with regard to model capabilities. </p><p>In November, Google announced the launch of Gemini 3. The new model range <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform"><u>outperformed OpenAI’s GPT-5 model</u></a> across a range of key benchmarks. That launch precipitated a <a href="https://www.theguardian.com/technology/2025/dec/02/sam-altman-issues-code-red-at-openai-as-chatgpt-contends-with-rivals"><u>similar ‘code red’ scramble at OpenAI</u></a>, with CEO Sam Altman pushing for the company to release GPT-5.2 in quick fashion. </p><h2 id="a-huge-victory-for-google">A huge victory for Google</h2><p>I wouldn’t be surprised if somewhere in the Googleplex, Alphabet’s sprawling Mountain View HQ, there isn’t a proud executive today hanging up a plaque that reads “Google's Al technology provides the most capable foundation for Apple Foundation Models”.</p><p>This is a major victory for Alphabet, particularly Google DeepMind and Google Cloud, which have spent the past year turning Google’s cloud empire into the foundations for AI dominance.</p><p>At Google Cloud Next 2025, I wrote that the firm had <a href="https://www.itpro.com/technology/artificial-intelligence/google-cloud-next-2025-enterprise-ai-adoption"><u>demonstrated its clear strength in enterprise AI</u></a>, backed by its colossal cloud infrastructure, <a href="https://www.itpro.com/infrastructure/why-google-cloud-is-betting-big-on-its-custom-chips"><u>home-grown TPUs</u></a> for AI training and inference, and first-hand access to the world’s data.</p><p>No doubt Alphabet will shout about this deal from the rafters at Google Cloud Next 2026.</p><p>Beyond the short-term victory laps, there are also significant long-term reputational gains for both Google and Apple that can arise from this deal.</p><p>After <a href="https://www.apple.com/uk/newsroom/2024/06/introducing-apple-intelligence-for-iphone-ipad-and-mac/" target="_blank"><u>first announcing</u></a> Apple Intelligence at WWDC 2024, Apple has had a rocky 18 months of repeated delays. Users who bought the iPhone 16 Pro on the understanding that they’d soon have a revamped Siri are still waiting for the promised updates.</p><p>In the meantime, the 16 Pro has been eclipsed by an entire generation of devices.</p><p>Apple has long practiced a ‘slow but effective’ approach to both its hardware and software, in which it shows up years late to technological trends but executes them to such a high standard that its user base can’t complain. Apple’s approach to AI has been no different, <a href="https://www.itpro.com/technology/artificial-intelligence/apples-generative-ai-slumber-is-ending-microsoft-and-google-should-be-on-alert"><u>as we’ve previously noted</u></a>. </p><p>With the iPhone, we’ve seen it with <a href="https://www.bbc.co.uk/news/technology-49636387#:~:text=The%20new%20iPhones%20are%20notable%20for%20introducing%20an%20%22ultrawide%22%20rear%20camera%2C%20offering%202x%20optical%20zoom%2Dout." target="_blank"><u>ultrawide lenses</u></a>, <a href="https://www.tomsguide.com/reviews/iphone-13-pro#:~:text=Of%20course%2C%20the,across%20the%20glass." target="_blank"><u>high refresh rate panels</u></a>, and later this year could see Apple <a href="https://www.techradar.com/news/foldable-iphone" target="_blank"><u>do it again with a foldable</u></a>.</p><p>But Apple users have no qualms about adopting competing software in the meantime – in the form of Gemini, OpenAI’s ChatGPT, and <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner"><u>Anthropic’s Claude</u></a>. Clearly, Apple has decided that its customers have waited too long and Google is the most reliable partner for immediate results.</p><p>Logan Kilpatrick, product lead for Google AI Studio and Gemini API, <a href="https://x.com/OfficialLoganK/status/1948158565581279672" target="_blank"><u>took to X</u></a> in July to announce that the firm had processed 980 trillion tokens in May, across its products and APIs. This number has since risen to <a href="https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q3-2025/#search:~:text=We%20are%20now%20processing%20over%201.3%20quadrillion%20monthly%20tokens" target="_blank">1.3 quadrillion monthly tokens</a>, with no notable outages across Google’s AI portfolio nor significant limitations imposed on free tier users.</p><p>Note also that this is a “multi-year” deal, with Google involved in crafting a range of new foundation models for Apple. Here the company can leverage its unparalleled access to data across Search, YouTube, Google Photos, and more, at a scale Apple can’t match.</p><p>The exact interdependencies between the Gemini family of models and Apple Intelligence is still to be determined, with Apple insisting the service will continue to operate both on device and using its <a href="https://www.itpro.com/security/apple-is-offering-rewards-of-up-to-usd1-million-to-find-critical-flaws-in-its-private-ai-cloud-systems">Private Cloud Compute</a> infrastructure.</p><p>In all likelihood, however, technical information will be kept at a minimum as Apple seeks to uphold a seamless AI experience for its iOS and macOS users.</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/googles-apple-deal-is-a-major-seal-of-approval-for-gemini-and-a-sure-sign-its-beginning-to-pull-ahead-of-openai-in-the-ai-race</link>
                                                                            <description>
                            <![CDATA[ Apple opting for Google's models to underpin Siri and Apple Intelligence is a major seal of approval for the tech giant's Gemini range – and a sure sign it's pulling ahead in the AI race. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">9aELk45shHdjJn6UhYerGG</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/dM84vmm4wJHX3c4gZCoPRX-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 13 Jan 2026 11:51:14 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:38 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                        <dc:contributor><![CDATA[ Rory Bathgate ]]></dc:contributor>
                                                                    <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/dM84vmm4wJHX3c4gZCoPRX-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Google CEO Sundar Pichai and Apple CEO Tim Cook pictured during a roundtable with American and Indian business leaders in the East Room of the White House on June 23, 2023.]]></media:description>                                                            <media:text><![CDATA[Google CEO Sundar Pichai and Apple CEO Tim Cook pictured during a roundtable with American and Indian business leaders in the East Room of the White House on June 23, 2023.]]></media:text>
                                <media:title type="plain"><![CDATA[Google CEO Sundar Pichai and Apple CEO Tim Cook pictured during a roundtable with American and Indian business leaders in the East Room of the White House on June 23, 2023.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/dM84vmm4wJHX3c4gZCoPRX-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Apple is turning to Google to help underpin its Siri service and “Apple Intelligence” capabilities, marking the latest sign the tech giant is gaining momentum in the AI race. </p><p>The duo have signed a multi-year partnership that will see Google support a major upgrade to Siri, which is expected to launch later this year. </p><p>Moving ahead, Apple’s foundation models will also be based on Google’s in-house Gemini AI model range, the company confirmed. </p><p>“After careful evaluation, Apple determined that Google's Al technology provides the most capable foundation for Apple Foundation Models and is excited about the innovative new experiences it will unlock for Apple users,” the duo said in a joint statement confirming the deal. </p><p>“Apple Intelligence will continue to run on Apple devices and Private Cloud Compute, while maintaining Apple's industry-leading privacy standards.”</p><p>The move comes in the wake of several months of speculation over closer ties between the two tech giants. Google already pays Apple billions of dollars to remain the default search engine on iPhones, for example. </p><p>According to reports from <a href="https://www.bloomberg.com/news/articles/2025-08-22/apple-explores-using-google-gemini-ai-to-power-revamped-siri?" target="_blank"><u><em>Bloomberg</em></u></a> in August last year, talks between the two over a potential deal for Gemini to power Siri were in the works. Separate <a href="https://www.bloomberg.com/news/articles/2025-11-05/apple-plans-to-use-1-2-trillion-parameter-google-gemini-model-to-power-new-siri" target="_blank"><u>reports </u></a>from the publication noted the deal could see Apple pay Google upwards of £1 billion a year to use the model range. </p><p>The deal also represents somewhat of a snub for OpenAI. Apple previously integrated ChatGPT into macOS and forged close ties with the company on Apple Intelligence, for example. </p><p>Opting for a deal with Google suggests the tech giant views the Gemini range as a more reliable option than OpenAI's flagship GPT series. </p><h2 id="google-is-pulling-ahead-in-the-ai-race">Google is pulling ahead in the AI race</h2><p>From Google’s perspective, the deal with Apple marks a huge seal of approval for its Gemini model range. Wedbush analyst Dan Ives said the partnership is a “major validation moment for Google” in a Monday research note, per reports from <a href="https://fortune.com/2026/01/12/apple-intelligence-powered-by-gemini-marks-a-major-validation-moment-for-google-top-tech-analyst-says/"><u><em>Fortune</em></u></a>. </p><p>Alphabet, Google’s parent company, hit a major milestone in the wake of the deal, surpassing a $4 trillion valuation for the first time. The surge means Alphabet is just the fourth company to hit such a milestone alongside Nvidia, Microsoft, and Apple. </p><p>A key factor behind the breakneck growth at Google has been its growth strength in the generative AI race. The company has recorded consistent growth and progress on this front over the last three years following a rocky start. </p><p>When <a href="https://www.itpro.com/technology/artificial-intelligence-ai/369965/what-is-chatgpt-and-what-does-it-mean-for-businesses">ChatGPT </a>launched in November 2022, Google was caught off guard by the release, prompting a rapid shift in focus led by CEO Sundar Pichai. </p><p>As <em>ITPro </em><a href="https://www.itpro.com/technology/artificial-intelligence-ai/369766/google-upends-teams-to-counter-threat-chatgpt"><u>reported at the time</u></a>, CEO Sundar Pichai issued a ‘code red’ alert which saw the company reallocate personnel and resources in a bid to keep pace with OpenAI and Microsoft. </p><p>Recent events suggest that Google has firmly caught up with key competitors in the AI space, particularly with regard to model capabilities. </p><p>In November, Google announced the launch of Gemini 3. The new model range <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform"><u>outperformed OpenAI’s GPT-5 model</u></a> across a range of key benchmarks. That launch precipitated a <a href="https://www.theguardian.com/technology/2025/dec/02/sam-altman-issues-code-red-at-openai-as-chatgpt-contends-with-rivals"><u>similar ‘code red’ scramble at OpenAI</u></a>, with CEO Sam Altman pushing for the company to release GPT-5.2 in quick fashion. </p><h2 id="a-huge-victory-for-google">A huge victory for Google</h2><p>I wouldn’t be surprised if somewhere in the Googleplex, Alphabet’s sprawling Mountain View HQ, there isn’t a proud executive today hanging up a plaque that reads “Google's Al technology provides the most capable foundation for Apple Foundation Models”.</p><p>This is a major victory for Alphabet, particularly Google DeepMind and Google Cloud, which have spent the past year turning Google’s cloud empire into the foundations for AI dominance.</p><p>At Google Cloud Next 2025, I wrote that the firm had <a href="https://www.itpro.com/technology/artificial-intelligence/google-cloud-next-2025-enterprise-ai-adoption"><u>demonstrated its clear strength in enterprise AI</u></a>, backed by its colossal cloud infrastructure, <a href="https://www.itpro.com/infrastructure/why-google-cloud-is-betting-big-on-its-custom-chips"><u>home-grown TPUs</u></a> for AI training and inference, and first-hand access to the world’s data.</p><p>No doubt Alphabet will shout about this deal from the rafters at Google Cloud Next 2026.</p><p>Beyond the short-term victory laps, there are also significant long-term reputational gains for both Google and Apple that can arise from this deal.</p><p>After <a href="https://www.apple.com/uk/newsroom/2024/06/introducing-apple-intelligence-for-iphone-ipad-and-mac/" target="_blank"><u>first announcing</u></a> Apple Intelligence at WWDC 2024, Apple has had a rocky 18 months of repeated delays. Users who bought the iPhone 16 Pro on the understanding that they’d soon have a revamped Siri are still waiting for the promised updates.</p><p>In the meantime, the 16 Pro has been eclipsed by an entire generation of devices.</p><p>Apple has long practiced a ‘slow but effective’ approach to both its hardware and software, in which it shows up years late to technological trends but executes them to such a high standard that its user base can’t complain. Apple’s approach to AI has been no different, <a href="https://www.itpro.com/technology/artificial-intelligence/apples-generative-ai-slumber-is-ending-microsoft-and-google-should-be-on-alert"><u>as we’ve previously noted</u></a>. </p><p>With the iPhone, we’ve seen it with <a href="https://www.bbc.co.uk/news/technology-49636387#:~:text=The%20new%20iPhones%20are%20notable%20for%20introducing%20an%20%22ultrawide%22%20rear%20camera%2C%20offering%202x%20optical%20zoom%2Dout." target="_blank"><u>ultrawide lenses</u></a>, <a href="https://www.tomsguide.com/reviews/iphone-13-pro#:~:text=Of%20course%2C%20the,across%20the%20glass." target="_blank"><u>high refresh rate panels</u></a>, and later this year could see Apple <a href="https://www.techradar.com/news/foldable-iphone" target="_blank"><u>do it again with a foldable</u></a>.</p><p>But Apple users have no qualms about adopting competing software in the meantime – in the form of Gemini, OpenAI’s ChatGPT, and <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner"><u>Anthropic’s Claude</u></a>. Clearly, Apple has decided that its customers have waited too long and Google is the most reliable partner for immediate results.</p><p>Logan Kilpatrick, product lead for Google AI Studio and Gemini API, <a href="https://x.com/OfficialLoganK/status/1948158565581279672" target="_blank"><u>took to X</u></a> in July to announce that the firm had processed 980 trillion tokens in May, across its products and APIs. This number has since risen to <a href="https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q3-2025/#search:~:text=We%20are%20now%20processing%20over%201.3%20quadrillion%20monthly%20tokens" target="_blank">1.3 quadrillion monthly tokens</a>, with no notable outages across Google’s AI portfolio nor significant limitations imposed on free tier users.</p><p>Note also that this is a “multi-year” deal, with Google involved in crafting a range of new foundation models for Apple. Here the company can leverage its unparalleled access to data across Search, YouTube, Google Photos, and more, at a scale Apple can’t match.</p><p>The exact interdependencies between the Gemini family of models and Apple Intelligence is still to be determined, with Apple insisting the service will continue to operate both on device and using its <a href="https://www.itpro.com/security/apple-is-offering-rewards-of-up-to-usd1-million-to-find-critical-flaws-in-its-private-ai-cloud-systems">Private Cloud Compute</a> infrastructure.</p><p>In all likelihood, however, technical information will be kept at a minimum as Apple seeks to uphold a seamless AI experience for its iOS and macOS users.</p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Microsoft CEO Satya Nadella wants an end to the term ‘AI slop’ in 2026 ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Microsoft CEO <a href="https://www.itpro.com/business/leadership/how-satya-nadellas-first-decade-at-microsoft-took-it-from-tech-giant-to-undisputed-industry-titan">Satya Nadella</a> believes 2026 will mark a “pivotal year” in the evolution of <a href="https://www.itpro.com/strategy/28181/what-is-ai">AI</a>, but the industry needs to shelve discussions over “slop” for real progress to be made. </p><p>In a <a href="https://snscratchpad.com/posts/looking-ahead-2026/" target="_blank"><u>blog post</u></a> before the turn of the new year, Nadella insisted businesses are making solid progress with integration of the technology, with many having “moved past the initial phase of discovery” and now entering a “phase of widespread diffusion”. </p><p>“As I reflect on the past year and look toward the one ahead, there’s no question 2026 will be a pivotal year for AI,” he wrote. </p><p>“We are beginning to distinguish between “spectacle” and “substance”. We now have a clearer sense of where the tech is headed, but also the harder and more important question of how to shape its impact on the world.”</p><p>Cutting through the marketing-style rhetoric here, Nadella suggested the technology is now at a point where enterprises are delivering value with AI, but does that all add up? Not entirely. </p><p>Return on investment (ROI) with AI tools is still a key hurdle for enterprise IT leaders, for example. A recent study from SAP noted that <a href="https://www.itpro.com/business/business-strategy/uk-firms-are-pouring-money-into-ai-but-they-wont-see-a-return-on-investment-unless-they-address-these-key-issues"><u>ROI is improving at a promising pace</u></a> but businesses are taking a fragmented approach to integration, creating longer-term challenges and impacting bottom lines. </p><p>This is an area Nadella pointed to in his 2025 roundup post, noting that he expects a key issue for IT leaders in the year ahead to be “model overhang”, whereby the capability of AI models and various tools outpaces the ability for teams to use it effectively.</p><p>Simply put, businesses aren’t fully maximizing their use of AI and they’re leaving some gas in the tank that could otherwise be used to fuel more progress. </p><p>To counter this, Nadella pointed to Steve Jobs’ famous “bicycles for the mind” metaphor as an example. Nadella said, AI should be used as a “scaffolding for human potential” in the same way Jobs described the ability for computers to enhance human capabilities. </p><p>“What matters is not the power of any given model, but how people choose to apply it to achieve their goals,” he explained. </p><h2 id="the-ai-slop-conundrum">The ‘AI slop’ conundrum</h2><p>Notably, Nadella insisted the industry needs to shelve discussions about “AI slop” if progress is to be made on unlocking full value from AI. </p><p>“We need to get beyond the arguments of slop vs sophistication and develop a new equilibrium in terms of our “theory of the mind” that accounts for humans being equipped with these new cognitive amplifier tools as we relate to each other,” he wrote.</p><p>“We will evolve from models to systems when it comes to deploying AI for real world impact.”</p><p>Describing AI as a “cognitive amplifier” does run counter to research by Microsoft last year which found the technology is having an <a href="https://www.itpro.com/technology/artificial-intelligence/ai-tools-critical-thinking-reliance"><u>adverse effect on cognitive function and critical thinking</u></a>. </p><p>Regardless, what this suggests is that IT leaders should focus on applying AI in instances where there are <a href="https://www.itpro.com/technology/artificial-intelligence/inside-aws-support-for-customer-ai-adoption-boots-on-the-ground">clear-cut use cases</a> for individual teams and needs – the technology isn’t a catch-all tool for use across the enterprise.</p><p>That’s easier said than done, however, particularly when one considers the hype surrounding the technology. Concerns over outputs and whether <a href="https://www.itpro.com/technology/artificial-intelligence/businesses-are-being-taken-for-fools-with-ai-agents">AI is delivering on the promises</a> made by big tech providers have become a recurring theme over the last year – particularly since the advent of <a href="https://www.itpro.com/technology/artificial-intelligence/practical-ai-the-age-of-agentic-ai">agentic AI</a>. </p><p>Microsoft, along with competitors such as <a href="https://www.itpro.com/cloud/infrastructure-as-a-service-iaas/362608/what-is-aws">Amazon Web Services (AWS)</a> and Google, have made significant strides with agentic AI across 2025, rolling out a range of services and new tools. Indeed, the tech giant’s annual Ignite<em> </em>conference in November was <a href="https://www.itpro.com/technology/artificial-intelligence/microsofts-new-agent-365-platform-is-a-one-stop-shop-for-deploying-securing-and-keeping-tabs-on-ai-agents">swamped with agentic AI product launches</a>. </p><p>But as <em>ITPro </em>reported in October, <a href="https://www.itpro.com/technology/artificial-intelligence/is-enterprise-agentic-ai-adoption-matching-the-hype"><u>agentic AI isn’t quite matching the hype</u></a>. OpenAI co-founder Andrej Karpathy poured cold water on the trend, describing the output of AI agents as “slop” and insisting the technology <a href="https://www.itpro.com/technology/artificial-intelligence/agentic-ai-hype-openai-andrej-karpathy"><u>isn’t living up to expectations</u></a>. </p><p>The issue of AI slop is already having an adverse impact on businesses across an array of industries. Open source project maintainers, for example, are encountering a huge surge in AI-generated “slop security reports”, as <a href="https://www.itpro.com/software/open-source/ai-slop-security-reports-are-driving-open-source-maintainers-mad"><u><em>ITPro </em></u><u>reported in late 2024</u></a>. </p><p>Elsewhere, a <a href="https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity" target="_blank"><u>study</u></a> published in <em>Harvard Business Review</em> last year found 40% of US employees encountered AI-generated “workslop” on a monthly basis. </p><p>This shoddy, AI-generated work content “masquerades as good work but lacks the substance to meaningfully advance a given task”, researchers noted. </p><p>Employees polled by researchers said around 15% of all work-related content they receive from colleagues falls under the “workslop” category. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/satya-nadella-microsoft-ai-slop-2026</link>
                                                                            <description>
                            <![CDATA[ Microsoft CEO Satya Nadella might want the term "AI slop" shelved in 2026, but businesses will still be dealing with increasing output problems and poor returns. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">ZwXdp6D74Sze9bsfTyAibG</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/h4sShHLFTwD854MYcjYCnT-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 05 Jan 2026 12:47:13 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:42 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/h4sShHLFTwD854MYcjYCnT-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Microsoft CEO Satya Nadella pictured speaking on stage during an event celebrating the 50th Anniversary of Microsoft on April 4, 2025 in Redmond, Washington.]]></media:description>                                                            <media:text><![CDATA[Microsoft CEO Satya Nadella pictured speaking on stage during an event celebrating the 50th Anniversary of Microsoft on April 4, 2025 in Redmond, Washington.]]></media:text>
                                <media:title type="plain"><![CDATA[Microsoft CEO Satya Nadella pictured speaking on stage during an event celebrating the 50th Anniversary of Microsoft on April 4, 2025 in Redmond, Washington.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/h4sShHLFTwD854MYcjYCnT-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Microsoft CEO <a href="https://www.itpro.com/business/leadership/how-satya-nadellas-first-decade-at-microsoft-took-it-from-tech-giant-to-undisputed-industry-titan">Satya Nadella</a> believes 2026 will mark a “pivotal year” in the evolution of <a href="https://www.itpro.com/strategy/28181/what-is-ai">AI</a>, but the industry needs to shelve discussions over “slop” for real progress to be made. </p><p>In a <a href="https://snscratchpad.com/posts/looking-ahead-2026/" target="_blank"><u>blog post</u></a> before the turn of the new year, Nadella insisted businesses are making solid progress with integration of the technology, with many having “moved past the initial phase of discovery” and now entering a “phase of widespread diffusion”. </p><p>“As I reflect on the past year and look toward the one ahead, there’s no question 2026 will be a pivotal year for AI,” he wrote. </p><p>“We are beginning to distinguish between “spectacle” and “substance”. We now have a clearer sense of where the tech is headed, but also the harder and more important question of how to shape its impact on the world.”</p><p>Cutting through the marketing-style rhetoric here, Nadella suggested the technology is now at a point where enterprises are delivering value with AI, but does that all add up? Not entirely. </p><p>Return on investment (ROI) with AI tools is still a key hurdle for enterprise IT leaders, for example. A recent study from SAP noted that <a href="https://www.itpro.com/business/business-strategy/uk-firms-are-pouring-money-into-ai-but-they-wont-see-a-return-on-investment-unless-they-address-these-key-issues"><u>ROI is improving at a promising pace</u></a> but businesses are taking a fragmented approach to integration, creating longer-term challenges and impacting bottom lines. </p><p>This is an area Nadella pointed to in his 2025 roundup post, noting that he expects a key issue for IT leaders in the year ahead to be “model overhang”, whereby the capability of AI models and various tools outpaces the ability for teams to use it effectively.</p><p>Simply put, businesses aren’t fully maximizing their use of AI and they’re leaving some gas in the tank that could otherwise be used to fuel more progress. </p><p>To counter this, Nadella pointed to Steve Jobs’ famous “bicycles for the mind” metaphor as an example. Nadella said, AI should be used as a “scaffolding for human potential” in the same way Jobs described the ability for computers to enhance human capabilities. </p><p>“What matters is not the power of any given model, but how people choose to apply it to achieve their goals,” he explained. </p><h2 id="the-ai-slop-conundrum">The ‘AI slop’ conundrum</h2><p>Notably, Nadella insisted the industry needs to shelve discussions about “AI slop” if progress is to be made on unlocking full value from AI. </p><p>“We need to get beyond the arguments of slop vs sophistication and develop a new equilibrium in terms of our “theory of the mind” that accounts for humans being equipped with these new cognitive amplifier tools as we relate to each other,” he wrote.</p><p>“We will evolve from models to systems when it comes to deploying AI for real world impact.”</p><p>Describing AI as a “cognitive amplifier” does run counter to research by Microsoft last year which found the technology is having an <a href="https://www.itpro.com/technology/artificial-intelligence/ai-tools-critical-thinking-reliance"><u>adverse effect on cognitive function and critical thinking</u></a>. </p><p>Regardless, what this suggests is that IT leaders should focus on applying AI in instances where there are <a href="https://www.itpro.com/technology/artificial-intelligence/inside-aws-support-for-customer-ai-adoption-boots-on-the-ground">clear-cut use cases</a> for individual teams and needs – the technology isn’t a catch-all tool for use across the enterprise.</p><p>That’s easier said than done, however, particularly when one considers the hype surrounding the technology. Concerns over outputs and whether <a href="https://www.itpro.com/technology/artificial-intelligence/businesses-are-being-taken-for-fools-with-ai-agents">AI is delivering on the promises</a> made by big tech providers have become a recurring theme over the last year – particularly since the advent of <a href="https://www.itpro.com/technology/artificial-intelligence/practical-ai-the-age-of-agentic-ai">agentic AI</a>. </p><p>Microsoft, along with competitors such as <a href="https://www.itpro.com/cloud/infrastructure-as-a-service-iaas/362608/what-is-aws">Amazon Web Services (AWS)</a> and Google, have made significant strides with agentic AI across 2025, rolling out a range of services and new tools. Indeed, the tech giant’s annual Ignite<em> </em>conference in November was <a href="https://www.itpro.com/technology/artificial-intelligence/microsofts-new-agent-365-platform-is-a-one-stop-shop-for-deploying-securing-and-keeping-tabs-on-ai-agents">swamped with agentic AI product launches</a>. </p><p>But as <em>ITPro </em>reported in October, <a href="https://www.itpro.com/technology/artificial-intelligence/is-enterprise-agentic-ai-adoption-matching-the-hype"><u>agentic AI isn’t quite matching the hype</u></a>. OpenAI co-founder Andrej Karpathy poured cold water on the trend, describing the output of AI agents as “slop” and insisting the technology <a href="https://www.itpro.com/technology/artificial-intelligence/agentic-ai-hype-openai-andrej-karpathy"><u>isn’t living up to expectations</u></a>. </p><p>The issue of AI slop is already having an adverse impact on businesses across an array of industries. Open source project maintainers, for example, are encountering a huge surge in AI-generated “slop security reports”, as <a href="https://www.itpro.com/software/open-source/ai-slop-security-reports-are-driving-open-source-maintainers-mad"><u><em>ITPro </em></u><u>reported in late 2024</u></a>. </p><p>Elsewhere, a <a href="https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity" target="_blank"><u>study</u></a> published in <em>Harvard Business Review</em> last year found 40% of US employees encountered AI-generated “workslop” on a monthly basis. </p><p>This shoddy, AI-generated work content “masquerades as good work but lacks the substance to meaningfully advance a given task”, researchers noted. </p><p>Employees polled by researchers said around 15% of all work-related content they receive from colleagues falls under the “workslop” category. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ 2025 marked the beginning of the end for OpenAI ]]></title>
                                                                                                <dc:content><![CDATA[ <p>OpenAI is closing out 2025 in a perilous position after what was a challenging year for the AI darling. </p><p>The company has basked in stardom since bursting onto the scene in late 2022. But eventually, the limelight fades and the rock solid foundations it’s built alongside Microsoft in the intervening years have taken a few major blows this year. </p><p>Indeed, 2025 started with a big hit: DeepSeek’s dramatic entrance to the global AI market. As <a href="https://www.itpro.com/technology/artificial-intelligence/chinese-ai-firm-deepseek-has-silicon-valley-flustered"><u><em>ITPro </em></u><u>noted at the time</u></a>, the launch of the company’s R1 <a href="https://www.itpro.com/technology/artificial-intelligence/generative-ai-vs-large-language-models">large language model (LLM)</a> rocked the US tech sector, wiping $1 trillion in stock value and prompting questions about the huge costs associated with AI training. </p><p>Adding insult to injury, within the space of a week, DeepSeek R1 <a href="https://techcrunch.com/2025/01/27/deepseek-displaces-chatgpt-as-the-app-stores-top-app/" target="_blank"><u>overtook ChatGPT on the US App Store</u></a>.</p><p>While it wasn’t quite the apocalyptic scenario some industry stakeholders suggested, the incident nonetheless marked the beginning of a year fraught with challenges – and ultimately OpenAI only had itself to blame. </p><p>The release of GPT-5 is a prime example here. The launch of the long-awaited AI model was <a href="https://www.itpro.com/technology/artificial-intelligence/openai-thought-it-hit-a-home-run-with-gpt-5-users-werent-so-keen"><u>slammed by users</u></a> over slow responses and basic errors. Even the tone of the model was criticized, resulting in calls for <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-openais-gpt-4o-updates-including-pricing-changes-and-new-features"><u>GPT-4o</u></a> to be brought back. </p><p>In the days after the launch, Sam Altman was fighting fires all over the place, <a href="https://x.com/sama/status/1953893841381273969" target="_blank"><u>taking to X</u></a> to insist the company “expected some bumpiness” with the roll-out and admitting it was “a little more bumpy than we hoped for”. </p><p>It’s safe to say that “bumpy” in this instance was an understatement. The incident highlighted a deeper problem for the company: it’s scrambling to release more powerful, exciting models to stay ahead of the competition – and it’s failing. </p><p>OpenAI is now embroiled in a war of attrition with key competitors such as Google and Anthropic. The latter’s models, such as <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner"><u>Claude Opus 4.5</u></a>, have improved at a rapid pace, particularly in notable areas such as coding capabilities. </p><p>Google, however, is the real threat for the company and November gave us a glimpse into the scale of the challenge OpenAI faces competing with the tech giant. <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform"><u>Gemini 3 blew GPT-5 out of the water</u></a>, outperforming the model across a range of key benchmarks. While writing this, Google doubled down with the launch of Gemini 3 Flash, which is quicker and cheaper than Gemini 3 <em>and </em>performs better on some benchmarks. </p><p>Compare the current situation to what went on three years ago, and the outlook isn’t good for OpenAI. When ChatGPT launched, the announcement reportedly sent Google into a meltdown behind the scenes. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence-ai/369766/google-upends-teams-to-counter-threat-chatgpt"><u>CEO Sundar Pichai initiated a “code red” alert</u></a>, reallocating staff and resources in a bid to push out its own AI tooling. The tech giant has been plugging away in the intervening years and that calm, collected approach appears to be paying dividends now. </p><p>By contrast, under the direction of Altman, OpenAI has been locked in its own code red scramble in recent weeks. Reports from the <a href="https://www.wsj.com/tech/ai/openais-altman-declares-code-red-to-improve-chatgpt-as-google-threatens-ai-lead-7faf5ea6?gaa_at=eafs&amp;gaa_n=AWEtsqdDWq-lRiBC7TXXlJgXAFNM3DNX_FC3hCwTpLRfoA3RtsXzE_At5ky3rSUXrCA%3D&amp;gaa_ts=6942cc2d&amp;gaa_sig=4f4jbqdH1-5EduHG1NM1gD-E5aFIlcHK8T5TmF-q76JR8FgtnnvH2FA9EvzEsaAAoPtdbJBbWWfA7lhyMKY2yw%3D%3D" target="_blank"><u><em>Wall Street Journal</em></u></a> in early December revealed Altman issued a call to action for developers, delaying development in some product areas to catch up with Google. </p><p>The resulting charging saw GPT-5.2 pushed out at breakneck speed, promising big performance improvements. That desperate scramble for OpenAI may have been for nothing, with MMMU-Pro benchmarks showing Gemini 3 Flash beats GPT-5.2 on multimodal reasoning with a score of 81.2% to 79.5%.</p><p>And so the arms race continues. </p><h2 id="rocky-ground-with-microsoft">Rocky ground with Microsoft</h2><p>OpenAI’s war of attrition is the least of its worries though. The company’s partnership with Microsoft - to the tune of several billion dollars in investment - was the fuel that propelled it to stardom in late 2022 and early 2023. But all signs point to <a href="https://www.itpro.com/technology/artificial-intelligence/the-honeymoon-period-is-officially-over-for-microsoft-and-openai"><u>a souring relationship with the tech giant</u></a>. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence/openai-chaos-what-next-for-sam-altman-and-the-beleaguered-tech-darling"><u>Sam Altman’s ousting</u></a> was the first big strain on the relationship in late 2023, and since then it appears things just haven’t been the same. Across 2025, the situation has been no different and Microsoft has been courting suitors elsewhere. </p><p>In early 2025, speculation about a <a href="https://www.reuters.com/technology/artificial-intelligence/microsoft-works-add-non-openai-models-into-365-copilot-products-sources-say-2024-12-23/"><u>big-money deal between Microsoft and Anthropic</u></a> to integrate the latter’s models within Microsoft 365 Copilot service hit the headlines. </p><p>At the time, <a href="https://www.itpro.com/software/microsoft/openai-continues-to-be-our-partner-on-frontier-models-microsoft-is-open-to-using-a-range-of-ai-models-in-365-copilot-but-openai-remains-its-go-to-choice"><u>Microsoft told </u><u><em>ITPro</em></u></a><em> </em>OpenAI “continues to be our partner on frontier models” and insisted the company was front and center in its future plans. </p><p>Fast forward to September, and <a href="https://www.microsoft.com/en-us/microsoft-365/blog/2025/09/24/expanding-model-choice-in-microsoft-365-copilot/"><u>Microsoft confirmed Claude will be available on the service</u></a>. This marked a huge blow for OpenAI, signalling Microsoft’s latest shift away from exclusivity. </p><p>On one front, OpenAI is being pushed to breaking point by Google on model development and release timelines. On the other, its longtime patron appears keen on distancing itself from the company. </p><h2 id="altman-bets-big-on-future-revenue">Altman bets big on future revenue</h2><p>2025 was the year OpenAI became synonymous with the idea of a looming “AI bubble”, largely fueled by its increasingly erratic and mind-boggling spending plans. </p><p>It’s no secret that the company needs a monumental amount of cash to keep running, and questions have been raised repeatedly <a href="https://www.itpro.com/technology/artificial-intelligence/openai-could-go-bankrupt-in-12-months-if-it-doesnt-raise-some-serious-cash-but-is-the-microsoft-backed-ai-giant-too-big-to-fail"><u>over its ability to basically stay afloat</u></a>.  This year alone the company has lined up more than $1 trillion in spending, signing megadeals <a href="https://www.itpro.com/cloud/cloud-computing/openai-oracle-cloud-deal-stargate"><u>with Oracle</u></a>, <a href="https://www.itpro.com/technology/artificial-intelligence/nvidia-openai-investment-100-billion"><u>Nvidia</u></a>, <a href="https://www.itpro.com/technology/artificial-intelligence/openai-signs-another-chip-deal-this-time-with-amd"><u>AMD</u></a>, Broadcom, and most recently <a href="https://www.itpro.com/cloud/cloud-computing/openai-just-signed-a-bumper-usd38bn-cloud-contract-with-aws-is-it-finally-preparing-to-cast-aside-microsoft"><u>with Amazon Web Services (AWS)</u></a>. </p><p>So how does it plan to pay for all this? Your guess is as good as mine. According to the <a href="https://www.ft.com/content/a169703c-c4df-46d6-a2d3-4184c74bbaf7" target="_blank"><u><em>Financial Times</em></u></a>, roughly 70% of the company’s $13bn in annual recurring revenue comes from the consumer version of ChatGPT. </p><p>The first objective here appears to be a significant increase in the number of paying subscribers. At present, that’s just 5% of the <a href="https://openai.com/index/the-state-of-enterprise-ai-2025-report/" target="_blank"><u>800 million users globally</u></a>. Setting this goal is one thing, actually getting them to fork out is another and seems wholly unrealistic. </p><p>There’s a sense that OpenAI is handcuffing itself to so many major industry players that if the worst comes to worse, it’ll either drag everyone into the abyss and brick the economy, or force said providers to take drastic measures to keep it in the game. </p><p>Reports from <a href="https://fortune.com/2025/10/07/data-centers-gdp-growth-zero-first-half-2025-jason-furman-harvard-economist/" target="_blank"><u><em>Fortune </em></u></a>in October this year noted that the lion’s share of US GDP growth in the first half of 2025 was fueled by investment in data centers infrastructure. We’re approaching a situation where if OpenAI goes kaput and drags everyone down with it, economic turmoil will follow. </p><h2 id="openai-has-no-clear-direction">OpenAI has no clear direction</h2><p>The clearest sign of OpenAI’s challenge across 2025 lies in the fact the company clearly doesn’t know exactly what it wants to be. </p><p>In May the company <a href="https://www.itpro.com/hardware/openai-just-made-its-first-hardware-acquisition-with-a-deal-for-jony-ives-startup-but-altmans-usd6-4-billion-moonshot-is-a-huge-risk"><u>acquired Jony Ive’s hardware firm</u></a>, io Products, and speculation over the launch of a “palm-sized personal assistant” next year has been growing, per reports from the <a href="https://www.ft.com/content/58b078be-e0ab-492f-9dbf-c2fe67298dd3"><u><em>Financial Times</em></u></a>. </p><p>This was the moment the penny dropped for me in 2025: OpenAI has no clear direction and is struggling to find its identity. It’s scrambling to appeal to consumers and enterprises alike with a myriad of paid tiers, grappling with huge compute costs and burning through cash, and has even suggested <a href="https://www.theverge.com/openai/648130/openai-social-network-x-competitor" target="_blank"><u>building a new social network</u></a>. </p><p>OpenAI has too many fingers in too many pies, and it’s rapidly running out of spare digits to stick elsewhere. </p> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/2025-marked-the-beginning-of-the-end-for-openai</link>
                                                                            <description>
                            <![CDATA[ OpenAI has its fingers in too many pies and it’s rapidly losing favor with consumers and enterprises alike ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">5d2xkMp6N92jmdnEodBipB</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/vsnrbm9y7aXXM8TEzC7Wpg-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Thu, 01 Jan 2026 08:30:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:47 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                <author><![CDATA[ ross.kelly@futurenet.com (Ross Kelly) ]]></author>                    <dc:creator><![CDATA[ Ross Kelly ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/Y5vrV2V98Np6jHAGmAtCd3-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Ross Kelly is ITPro&#039;s News &amp;amp; Analysis Editor, with a keen interest in cyber security, business leadership and emerging technologies.&lt;/p&gt;
&lt;p&gt;He graduated from Edinburgh Napier University in 2016 with a BA (Hons) in Journalism, and joined ITPro in 2022 after four years working in technology conference research.&amp;nbsp;&lt;/p&gt;
&lt;p&gt;In his spare time, Ross enjoys cycling, walking and is an avid reader of history and non-fiction.&lt;/p&gt;
&lt;p&gt;You can contact Ross at ross.kelly@futurenet.com or on &lt;a href=&quot;https://twitter.com/rosswritesetc&quot;&gt;Twitter&lt;/a&gt; and &lt;a href=&quot;https://www.linkedin.com/in/ross-kelly-18a54411a/&quot;&gt;LinkedIn&lt;/a&gt;.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/vsnrbm9y7aXXM8TEzC7Wpg-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A black and white photo of Sam Altman, CEO at OpenAI, on a blue background with a phone bearing the OpenAI logo held in the foreground.]]></media:description>                                                            <media:text><![CDATA[A black and white photo of Sam Altman, CEO at OpenAI, on a blue background with a phone bearing the OpenAI logo held in the foreground.]]></media:text>
                                <media:title type="plain"><![CDATA[A black and white photo of Sam Altman, CEO at OpenAI, on a blue background with a phone bearing the OpenAI logo held in the foreground.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/vsnrbm9y7aXXM8TEzC7Wpg-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>OpenAI is closing out 2025 in a perilous position after what was a challenging year for the AI darling. </p><p>The company has basked in stardom since bursting onto the scene in late 2022. But eventually, the limelight fades and the rock solid foundations it’s built alongside Microsoft in the intervening years have taken a few major blows this year. </p><p>Indeed, 2025 started with a big hit: DeepSeek’s dramatic entrance to the global AI market. As <a href="https://www.itpro.com/technology/artificial-intelligence/chinese-ai-firm-deepseek-has-silicon-valley-flustered"><u><em>ITPro </em></u><u>noted at the time</u></a>, the launch of the company’s R1 <a href="https://www.itpro.com/technology/artificial-intelligence/generative-ai-vs-large-language-models">large language model (LLM)</a> rocked the US tech sector, wiping $1 trillion in stock value and prompting questions about the huge costs associated with AI training. </p><p>Adding insult to injury, within the space of a week, DeepSeek R1 <a href="https://techcrunch.com/2025/01/27/deepseek-displaces-chatgpt-as-the-app-stores-top-app/" target="_blank"><u>overtook ChatGPT on the US App Store</u></a>.</p><p>While it wasn’t quite the apocalyptic scenario some industry stakeholders suggested, the incident nonetheless marked the beginning of a year fraught with challenges – and ultimately OpenAI only had itself to blame. </p><p>The release of GPT-5 is a prime example here. The launch of the long-awaited AI model was <a href="https://www.itpro.com/technology/artificial-intelligence/openai-thought-it-hit-a-home-run-with-gpt-5-users-werent-so-keen"><u>slammed by users</u></a> over slow responses and basic errors. Even the tone of the model was criticized, resulting in calls for <a href="https://www.itpro.com/technology/artificial-intelligence/everything-you-need-to-know-about-openais-gpt-4o-updates-including-pricing-changes-and-new-features"><u>GPT-4o</u></a> to be brought back. </p><p>In the days after the launch, Sam Altman was fighting fires all over the place, <a href="https://x.com/sama/status/1953893841381273969" target="_blank"><u>taking to X</u></a> to insist the company “expected some bumpiness” with the roll-out and admitting it was “a little more bumpy than we hoped for”. </p><p>It’s safe to say that “bumpy” in this instance was an understatement. The incident highlighted a deeper problem for the company: it’s scrambling to release more powerful, exciting models to stay ahead of the competition – and it’s failing. </p><p>OpenAI is now embroiled in a war of attrition with key competitors such as Google and Anthropic. The latter’s models, such as <a href="https://www.itpro.com/technology/artificial-intelligence/anthropic-announces-claude-opus-4-5-the-new-ai-coding-frontrunner"><u>Claude Opus 4.5</u></a>, have improved at a rapid pace, particularly in notable areas such as coding capabilities. </p><p>Google, however, is the real threat for the company and November gave us a glimpse into the scale of the challenge OpenAI faces competing with the tech giant. <a href="https://www.itpro.com/technology/artificial-intelligence/google-launches-flagship-gemini-3-model-and-google-antigravity-a-new-agentic-ai-development-platform"><u>Gemini 3 blew GPT-5 out of the water</u></a>, outperforming the model across a range of key benchmarks. While writing this, Google doubled down with the launch of Gemini 3 Flash, which is quicker and cheaper than Gemini 3 <em>and </em>performs better on some benchmarks. </p><p>Compare the current situation to what went on three years ago, and the outlook isn’t good for OpenAI. When ChatGPT launched, the announcement reportedly sent Google into a meltdown behind the scenes. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence-ai/369766/google-upends-teams-to-counter-threat-chatgpt"><u>CEO Sundar Pichai initiated a “code red” alert</u></a>, reallocating staff and resources in a bid to push out its own AI tooling. The tech giant has been plugging away in the intervening years and that calm, collected approach appears to be paying dividends now. </p><p>By contrast, under the direction of Altman, OpenAI has been locked in its own code red scramble in recent weeks. Reports from the <a href="https://www.wsj.com/tech/ai/openais-altman-declares-code-red-to-improve-chatgpt-as-google-threatens-ai-lead-7faf5ea6?gaa_at=eafs&amp;gaa_n=AWEtsqdDWq-lRiBC7TXXlJgXAFNM3DNX_FC3hCwTpLRfoA3RtsXzE_At5ky3rSUXrCA%3D&amp;gaa_ts=6942cc2d&amp;gaa_sig=4f4jbqdH1-5EduHG1NM1gD-E5aFIlcHK8T5TmF-q76JR8FgtnnvH2FA9EvzEsaAAoPtdbJBbWWfA7lhyMKY2yw%3D%3D" target="_blank"><u><em>Wall Street Journal</em></u></a> in early December revealed Altman issued a call to action for developers, delaying development in some product areas to catch up with Google. </p><p>The resulting charging saw GPT-5.2 pushed out at breakneck speed, promising big performance improvements. That desperate scramble for OpenAI may have been for nothing, with MMMU-Pro benchmarks showing Gemini 3 Flash beats GPT-5.2 on multimodal reasoning with a score of 81.2% to 79.5%.</p><p>And so the arms race continues. </p><h2 id="rocky-ground-with-microsoft">Rocky ground with Microsoft</h2><p>OpenAI’s war of attrition is the least of its worries though. The company’s partnership with Microsoft - to the tune of several billion dollars in investment - was the fuel that propelled it to stardom in late 2022 and early 2023. But all signs point to <a href="https://www.itpro.com/technology/artificial-intelligence/the-honeymoon-period-is-officially-over-for-microsoft-and-openai"><u>a souring relationship with the tech giant</u></a>. </p><p><a href="https://www.itpro.com/technology/artificial-intelligence/openai-chaos-what-next-for-sam-altman-and-the-beleaguered-tech-darling"><u>Sam Altman’s ousting</u></a> was the first big strain on the relationship in late 2023, and since then it appears things just haven’t been the same. Across 2025, the situation has been no different and Microsoft has been courting suitors elsewhere. </p><p>In early 2025, speculation about a <a href="https://www.reuters.com/technology/artificial-intelligence/microsoft-works-add-non-openai-models-into-365-copilot-products-sources-say-2024-12-23/"><u>big-money deal between Microsoft and Anthropic</u></a> to integrate the latter’s models within Microsoft 365 Copilot service hit the headlines. </p><p>At the time, <a href="https://www.itpro.com/software/microsoft/openai-continues-to-be-our-partner-on-frontier-models-microsoft-is-open-to-using-a-range-of-ai-models-in-365-copilot-but-openai-remains-its-go-to-choice"><u>Microsoft told </u><u><em>ITPro</em></u></a><em> </em>OpenAI “continues to be our partner on frontier models” and insisted the company was front and center in its future plans. </p><p>Fast forward to September, and <a href="https://www.microsoft.com/en-us/microsoft-365/blog/2025/09/24/expanding-model-choice-in-microsoft-365-copilot/"><u>Microsoft confirmed Claude will be available on the service</u></a>. This marked a huge blow for OpenAI, signalling Microsoft’s latest shift away from exclusivity. </p><p>On one front, OpenAI is being pushed to breaking point by Google on model development and release timelines. On the other, its longtime patron appears keen on distancing itself from the company. </p><h2 id="altman-bets-big-on-future-revenue">Altman bets big on future revenue</h2><p>2025 was the year OpenAI became synonymous with the idea of a looming “AI bubble”, largely fueled by its increasingly erratic and mind-boggling spending plans. </p><p>It’s no secret that the company needs a monumental amount of cash to keep running, and questions have been raised repeatedly <a href="https://www.itpro.com/technology/artificial-intelligence/openai-could-go-bankrupt-in-12-months-if-it-doesnt-raise-some-serious-cash-but-is-the-microsoft-backed-ai-giant-too-big-to-fail"><u>over its ability to basically stay afloat</u></a>.  This year alone the company has lined up more than $1 trillion in spending, signing megadeals <a href="https://www.itpro.com/cloud/cloud-computing/openai-oracle-cloud-deal-stargate"><u>with Oracle</u></a>, <a href="https://www.itpro.com/technology/artificial-intelligence/nvidia-openai-investment-100-billion"><u>Nvidia</u></a>, <a href="https://www.itpro.com/technology/artificial-intelligence/openai-signs-another-chip-deal-this-time-with-amd"><u>AMD</u></a>, Broadcom, and most recently <a href="https://www.itpro.com/cloud/cloud-computing/openai-just-signed-a-bumper-usd38bn-cloud-contract-with-aws-is-it-finally-preparing-to-cast-aside-microsoft"><u>with Amazon Web Services (AWS)</u></a>. </p><p>So how does it plan to pay for all this? Your guess is as good as mine. According to the <a href="https://www.ft.com/content/a169703c-c4df-46d6-a2d3-4184c74bbaf7" target="_blank"><u><em>Financial Times</em></u></a>, roughly 70% of the company’s $13bn in annual recurring revenue comes from the consumer version of ChatGPT. </p><p>The first objective here appears to be a significant increase in the number of paying subscribers. At present, that’s just 5% of the <a href="https://openai.com/index/the-state-of-enterprise-ai-2025-report/" target="_blank"><u>800 million users globally</u></a>. Setting this goal is one thing, actually getting them to fork out is another and seems wholly unrealistic. </p><p>There’s a sense that OpenAI is handcuffing itself to so many major industry players that if the worst comes to worse, it’ll either drag everyone into the abyss and brick the economy, or force said providers to take drastic measures to keep it in the game. </p><p>Reports from <a href="https://fortune.com/2025/10/07/data-centers-gdp-growth-zero-first-half-2025-jason-furman-harvard-economist/" target="_blank"><u><em>Fortune </em></u></a>in October this year noted that the lion’s share of US GDP growth in the first half of 2025 was fueled by investment in data centers infrastructure. We’re approaching a situation where if OpenAI goes kaput and drags everyone down with it, economic turmoil will follow. </p><h2 id="openai-has-no-clear-direction">OpenAI has no clear direction</h2><p>The clearest sign of OpenAI’s challenge across 2025 lies in the fact the company clearly doesn’t know exactly what it wants to be. </p><p>In May the company <a href="https://www.itpro.com/hardware/openai-just-made-its-first-hardware-acquisition-with-a-deal-for-jony-ives-startup-but-altmans-usd6-4-billion-moonshot-is-a-huge-risk"><u>acquired Jony Ive’s hardware firm</u></a>, io Products, and speculation over the launch of a “palm-sized personal assistant” next year has been growing, per reports from the <a href="https://www.ft.com/content/58b078be-e0ab-492f-9dbf-c2fe67298dd3"><u><em>Financial Times</em></u></a>. </p><p>This was the moment the penny dropped for me in 2025: OpenAI has no clear direction and is struggling to find its identity. It’s scrambling to appeal to consumers and enterprises alike with a myriad of paid tiers, grappling with huge compute costs and burning through cash, and has even suggested <a href="https://www.theverge.com/openai/648130/openai-social-network-x-competitor" target="_blank"><u>building a new social network</u></a>. </p><p>OpenAI has too many fingers in too many pies, and it’s rapidly running out of spare digits to stick elsewhere. </p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ OpenAI says prompt injection attacks are a serious threat for AI browsers ]]></title>
                                                                                                <dc:content><![CDATA[ <p>OpenAI has updated its browser to boost protection against prompt injection attacks, but it warned the risk may never fully disappear. </p><p>Released in October, OpenAI's ChatGPT Atlas browser includes agent mode, which looks at webpages to click its way through transactions, forms and other online tasks.</p><p>But OpenAI noted that as a browser agent can do more, it also becomes more at risk to "adversarial attacks" – in particular prompt injections, which are sneaking malicious instructions into the agent to drive its behavior. </p><p>"Prompt injection⁠ is one of the most significant risks we actively defend against to help ensure ChatGPT Atlas can operate securely on your behalf," OpenAI said in a <a href="https://openai.com/index/hardening-atlas-against-prompt-injection/" target="_blank"><u>blog post</u></a>.</p><p>Indeed, <a href="https://www.itpro.com/security/cyber-researchers-have-already-identified-several-big-security-vulnerabilities-on-openais-atlas-browser"><u>days after the release of OpenAI's browser</u></a>, security researchers spotted several serious flaws, including a prompt-injection technique – no wonder then that analysts at <a href="https://www.theregister.com/2025/12/08/gartner_recommends_ai_browser_ban/"><u>Gartner have warned</u></a> companies to ban AI browsers for fear of security risks.</p><p>OpenAI said it recently updated ChatGPT Atlas's agent security safeguards and gave it a new model that had been "adversarially trained", as well developing a "rapid response loop" to find flaws and address them. </p><p>That was sparked by red teaming, in which an internal team acts like threat actors to test the system for flaws or weaknesses. In this instance, what they found suggests prompt injection is a "long-term AI security challenge". </p><p>"<a href="https://www.itpro.com/security/ncsc-issues-urgent-warning-over-growing-ai-prompt-injection-risks-heres-what-you-need-to-know">Prompt injection</a>, much like scams and social engineering on the web, is unlikely to ever be fully 'solved'," the company said.</p><p>"But we’re optimistic that a proactive, highly responsive rapid response  loop can continue to materially reduce real-world risk over time," OpenAI added. </p><p>"By combining automated attack discovery with adversarial training and system-level safeguards, we can identify new attack patterns earlier, close gaps faster, and continuously raise the cost of exploitation."</p><h2 id="new-challenge-for-ai-browsers">New challenge for AI browsers</h2><p>Prompt injection is when attackers get between an agent's prompt box and the AI model, changing instructions to create malicious results. It's a new problem for browsers that boast AI features – and there’s a growing array already. </p><p>The main issue here is that since agents can take many of the same actions as a user, OpenAI said the potential impact of a successful attack could be “just as broad”. </p><p>As an example, OpenAI said an attacker could send a malicious email to trick an agent to ignore the user's actual request in favour of forwarding sensitive documents. </p><p>The user gives access to email for a legitimate task, such as summarizing messages, but if the agent also follows the injected instructions, sensitive data could leak.</p><h2 id="fighting-back">Fighting back</h2><p>OpenAI has made previous efforts to protect against such attacks, but is now adding new techniques to help avoid prompt injections. </p><p>First, the company has built an AI-powered hacker to use as an automated red teaming tool to proactively hunt out prompt injection attacks – even complicated ones taking hundreds of steps. </p><p>"We trained this attacker end-to-end with reinforcement learning, so it learns from its own successes and failures to improve its red teaming skills," the company said. </p><p>Beyond that, OpenAI has developed what it calls a "rapid response loop". When that automated <a href="https://www.itpro.com/security/34590/stories-from-the-front-line-the-secrets-of-the-red-team-revealed">red team</a> spots a potential injection technique, that's fed back into the AI via adversarial training.</p><p>"We continuously train updated agent models against our best automated attacker—prioritizing the attacks where the target agents currently fail," the company added. </p><p>"The goal is to teach agents to ignore adversarial instructions and stay aligned with the user’s intent, improving resistance to newly discovered prompt-injection strategies."</p><p>Similarly, when using the agent in the ChatGPT Atlas browser, OpenAI advised using "logged out" mode where possible, signing in only when necessary to complete a task, and told users to review all confirmation requests carefully. </p><p>When it comes to prompts, be specific rather than broad: saying "review my emails and take whatever action is needed" gives space for threat actors to meddle. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3> ]]></dc:content>
                                                                                                                                            <link>https://www.itpro.com/technology/artificial-intelligence/openai-chatgpt-atlas-ai-browser-prompt-injection-attack-risk</link>
                                                                            <description>
                            <![CDATA[ OpenAI details efforts to protect ChatGPT Atlas against prompt injection attacks ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">vEWEZvJuJb4vDev2GnEP</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/khbzGP2pUAektJZSKgYaHF-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Wed, 24 Dec 2025 09:00:00 +0000</pubDate>                                                                                                                                <updated>Thu, 19 Nov 2026 12:34:17 +0000</updated>
                                                                                                                                            <category><![CDATA[Generative AI]]></category>
                                                    <category><![CDATA[Artificial Intelligence]]></category>
                                                                                                                    <dc:creator><![CDATA[ Nicole Kobie ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/8Y8JDDTQ7XDEk49FoAFP2S-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Nicole Kobie first started writing for ITPro in 2007. As a freelance journalist covering technology and business, Nicole&#039;s work includes  bylines in New Scientist, Wired, PC Pro and many more. &lt;/p&gt;&lt;p&gt;Nicole the author of a book about the history of technology, The Long History of the Future.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/khbzGP2pUAektJZSKgYaHF-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Logo of OpenAI, developer of the GPT-4.1 AI model family, pictured on a smartphone screen placed on a table.]]></media:description>                                                            <media:text><![CDATA[Logo of OpenAI, developer of the GPT-4.1 AI model family, pictured on a smartphone screen placed on a table.]]></media:text>
                                <media:title type="plain"><![CDATA[Logo of OpenAI, developer of the GPT-4.1 AI model family, pictured on a smartphone screen placed on a table.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/khbzGP2pUAektJZSKgYaHF-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>OpenAI has updated its browser to boost protection against prompt injection attacks, but it warned the risk may never fully disappear. </p><p>Released in October, OpenAI's ChatGPT Atlas browser includes agent mode, which looks at webpages to click its way through transactions, forms and other online tasks.</p><p>But OpenAI noted that as a browser agent can do more, it also becomes more at risk to "adversarial attacks" – in particular prompt injections, which are sneaking malicious instructions into the agent to drive its behavior. </p><p>"Prompt injection⁠ is one of the most significant risks we actively defend against to help ensure ChatGPT Atlas can operate securely on your behalf," OpenAI said in a <a href="https://openai.com/index/hardening-atlas-against-prompt-injection/" target="_blank"><u>blog post</u></a>.</p><p>Indeed, <a href="https://www.itpro.com/security/cyber-researchers-have-already-identified-several-big-security-vulnerabilities-on-openais-atlas-browser"><u>days after the release of OpenAI's browser</u></a>, security researchers spotted several serious flaws, including a prompt-injection technique – no wonder then that analysts at <a href="https://www.theregister.com/2025/12/08/gartner_recommends_ai_browser_ban/"><u>Gartner have warned</u></a> companies to ban AI browsers for fear of security risks.</p><p>OpenAI said it recently updated ChatGPT Atlas's agent security safeguards and gave it a new model that had been "adversarially trained", as well developing a "rapid response loop" to find flaws and address them. </p><p>That was sparked by red teaming, in which an internal team acts like threat actors to test the system for flaws or weaknesses. In this instance, what they found suggests prompt injection is a "long-term AI security challenge". </p><p>"<a href="https://www.itpro.com/security/ncsc-issues-urgent-warning-over-growing-ai-prompt-injection-risks-heres-what-you-need-to-know">Prompt injection</a>, much like scams and social engineering on the web, is unlikely to ever be fully 'solved'," the company said.</p><p>"But we’re optimistic that a proactive, highly responsive rapid response  loop can continue to materially reduce real-world risk over time," OpenAI added. </p><p>"By combining automated attack discovery with adversarial training and system-level safeguards, we can identify new attack patterns earlier, close gaps faster, and continuously raise the cost of exploitation."</p><h2 id="new-challenge-for-ai-browsers">New challenge for AI browsers</h2><p>Prompt injection is when attackers get between an agent's prompt box and the AI model, changing instructions to create malicious results. It's a new problem for browsers that boast AI features – and there’s a growing array already. </p><p>The main issue here is that since agents can take many of the same actions as a user, OpenAI said the potential impact of a successful attack could be “just as broad”. </p><p>As an example, OpenAI said an attacker could send a malicious email to trick an agent to ignore the user's actual request in favour of forwarding sensitive documents. </p><p>The user gives access to email for a legitimate task, such as summarizing messages, but if the agent also follows the injected instructions, sensitive data could leak.</p><h2 id="fighting-back">Fighting back</h2><p>OpenAI has made previous efforts to protect against such attacks, but is now adding new techniques to help avoid prompt injections. </p><p>First, the company has built an AI-powered hacker to use as an automated red teaming tool to proactively hunt out prompt injection attacks – even complicated ones taking hundreds of steps. </p><p>"We trained this attacker end-to-end with reinforcement learning, so it learns from its own successes and failures to improve its red teaming skills," the company said. </p><p>Beyond that, OpenAI has developed what it calls a "rapid response loop". When that automated <a href="https://www.itpro.com/security/34590/stories-from-the-front-line-the-secrets-of-the-red-team-revealed">red team</a> spots a potential injection technique, that's fed back into the AI via adversarial training.</p><p>"We continuously train updated agent models against our best automated attacker—prioritizing the attacks where the target agents currently fail," the company added. </p><p>"The goal is to teach agents to ignore adversarial instructions and stay aligned with the user’s intent, improving resistance to newly discovered prompt-injection strategies."</p><p>Similarly, when using the agent in the ChatGPT Atlas browser, OpenAI advised using "logged out" mode where possible, signing in only when necessary to complete a task, and told users to review all confirmation requests carefully. </p><p>When it comes to prompts, be specific rather than broad: saying "review my emails and take whatever action is needed" gives space for threat actors to meddle. </p><h3 class="article-body__section" id="section-follow-us-on-social-media"><span>FOLLOW US ON SOCIAL MEDIA</span></h3>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
            </channel>
</rss>