Google Cloud announces UK data residency for agentic AI services

Google Cloud has announced new sovereign services in the UK, as well as new skills and cloud services support for UK AI startups and established British businesses.

In a commitment to privacy-focused and restricted UK businesses, Google cloud will offer UK data residency for Google Agentspace, its platform for enterprise AI agents. This will allow organizations to use multimodal AI searches and AI agents tailored to their companies, while keeping data securely stored within the UK.

Google Cloud stated that the feature would be available to UK customers starting in its Q2, which runs from April to June.

The announcement is an expansion of Google Cloud’s UK data residency offerings, with the cloud giant having previously affirmed support for storing at-rest Vertex AI data in the UK alongside UK-limited machine learning processing for Gemini 1.5 Flash.

Google Cloud made the announcements at the Gemini for the United Kingdom live event, held at the Google DeepMind headquarters in London, UK.

Data sovereignty has become a priority for UK organizations, amid a wider push for sovereign cloud capabilities across the world.

Allison Kirkby, CEO at BT Group, hailed the announcement as an important step to improving customer trust and reducing the concerns of UK businesses looking to use AI.

"Our overall mission as a company is to be the most trusted connector of people, business and society," said Kirkby.

"And considering the world we live in, we certainly have customers that want to ensure that their data is safe and secure and its on platforms that are encrypted in the right way, with the right people handling it."

In addition to its expanded UK cloud offering, Google Cloud also announced new support packages for British business.

At the top level, Google Cloud announced £280,000 in cloud credits for UK AI startups, to be delivered in conjunction with expert-led training, mentorships, and team workshops. Cloud credits can be spent on all Google Cloud services, to leverage them at no cost.

Over 60% of all UK generative AI startups already use Google Cloud, per the company’s internal stats. The announcement will see it further expand support for the country’s burgeoning tech startup scene.

"Our deep relationship with Google DeepMind enables us to bring some of the world’s most cutting-edge AI technology to our cloud customers, which range from startups to large enterprises," said Thomas Kurian, CEO, Google Cloud.

“We're uniquely able to provide secure, flexible infrastructure; leading AI models; and an open developer platform that integrates with existing IT investments while maintaining security, privacy, and access controls, to organisations in the UK and around the world.”

Google Cloud ramps up training schemes

Google Cloud will also offer new training programs for professionals and students to improve digital literacy and the UK’s AI skills.

The firm will offer UK-based developers 35 credits per month to use on Google Cloud Skills Boost, its online learning platform, to improve their knowledge of using generative AI in the workplace. Alongside this, students will be given access to the Google Career Launchpad, which opens them to free, entry-level Google Cloud certifications in data analytics and cybersecurity skills.

Alongside its skills commitments, Google Cloud pointed to its January 2024 announcement of a $1 billion (£770 million) data center in Hertfordshire as an example of its ongoing investment in the UK. The site is one of several high profile data centers announced in the past year, as the UK seeks to meet surging data center demand.

The company continues to expand its data center presence as it rolls out Gemini across Google Workspace and widens its AI offerings available via Vertex AI.

Another was announced at the event, in the form of Google Cloud HD Voices, a new synthesized voice capability on Vertex AI. The service, powered by Google Cloud’s next generation text-to-speech model Chirp 3, promises to provide businesses with 248 lifelike voices across 31 languages.

Google continues to limit complete access to Chirp 3, which can be trained to replicate any voice given sufficient training, over concerns that it could be used to maliciously clone voices. Only organizations that have been manually cleared by Google Cloud have access to the model itself.

Responding to an ITPro question on when the voice cloning features of Chirp 3 are likely to become available to the general public, Kurian stated that Google Cloud will release the model when it meets its own safety controls and gets approval from its trusted partners.

“Whenever we introduce a new modality, we typically take a few months with customers who are in preview, testing the kinds of safety concerns people have, for example is it using abusive language and in the case of Chirp, can you teach it to respond empathetically so that the voice sounds empathetic."

Kurian pointed to other models to give examples of Google Cloud staggering the release of products in the name of safety.

“When we launched Imagen, we went through a similar process, when we launched Veo, we went through a similar process, it's just during the preview phase we work with a few customers to get feedback and the developer community. And once we have that, we typically fold that in and make the product generally available, including the safety controls around it.”

MORE FROM ITPRO

• What is a sovereign cloud?
• Google Cloud announces data residency for machine learning in the UK
• New GitHub rules mean users can store code and repository data in the EU