AI-generated code is now the cause of one-in-five breaches – but developers and security leaders alike are convinced the technology will come good eventually
Most security leaders still think AI tools will eventually write secure, reliable, code
AI coding tools are creating serious security risks in production, with one-in-five CISOs saying they've suffered major incidents because of AI-generated code.
AI coding tools now write 24% of production code – 21% in Europe and 29% in the US – according to a new report from Aikido. But it's risky, with 69% of security leaders, security engineers, and developers across Europe and the US revealing they'd found serious vulnerabilities in AI-written code.
US-based respondents were among the worst hit by AI-related flaws, with 43% of organizations reporting serious incidents, compared with just 20% in Europe.
This, the study noted, appears to be down to better prevention and oversight. For example, EU-based firms reported more “near misses” with AI-generated code than their US counterparts, potentially highlighting more robust testing practices.
Adding more tools to address the issue isn’t helping, Aikido found. Indeed, organizations with more security tools report more incidents, with more overhead and slower remediation.
Nearly two-thirds (64%) of those with just one or two tools had an incident, the figure was 90% for those with between six and nine tools.
All-in-one AI coding tools are helping bridge gaps
Notably, teams using tools designed for both developers and security teams were more than twice as likely to report zero incidents than those using tools made for only one specific group.
Sign up today and you will receive a free copy of our Future Focus 2026 report - the leading resource for IT decision-maker insight on priorities and investment areas in AI, security and more.
“Giving developers the right security tool that works with existing tools and workflows allows teams to implement security best practices and improve their posture,” commented Walid Mahmoud, DevSecOps lead at the UK Cabinet Office.
Teams using separate AppSec and CloudSec tools were 50% more likely to face incidents, and 93% of those with separate tools reported integration headaches such as duplicate alerts or inconsistent data.
The security blame game is heating up
The blame for incidents caused by AI code is now becoming a serious point of contention within enterprises, the report noted. For example, 53% of respondents blamed security teams for failing to address issues, while 45% blamed developers who failed to spot issues before pushing to production.
Meanwhile, 42% pointed toward whoever merged it. This blame game is expected to continue escalating, according to Aikido. Half of developers reckoned they’d be blamed if the AI code they wrote introduced a vulnerability, even more than the security team itself.
“There's clearly a lack of clarity among respondents over where accountability should sit for good risk management,” commented Andy Boura, CISO at Rothesay.
Despite concerns across the board, enterprises are expected to continue driving ahead with adoption of AI coding tools, the study noted. Nine-in-ten said they expect AI to take over penetration testing within the next five years, for example
Meanwhile, 96% believe AI will write secure, reliable, code at some point, with the biggest proportion (44%) thinking it will happen in the next three-to-five years.
Only 21% think this will be achieved without human oversight, however, underlining the importance of keeping humans in the loop.
Make sure to follow ITPro on Google News to keep tabs on all our latest news, analysis, and reviews.
MORE FROM ITPRO
- Think AI coding tools are speeding up work? Think again – they’re actually slowing developers down
- How AI coding is transforming the IT industry in 2025
- AI coding tools are booming – and developers in this one country are by far the most frequent users
Emma Woollacott is a freelance journalist writing for publications including the BBC, Private Eye, Forbes, Raconteur and specialist technology titles.
-
TfL takes Oyster photocard systems offline following breachNews Security experts warn the Oyster photocard breach could lead to follow-up attacks on victims
-
Irish credit unions hit by data breach after IT provider compromisedNews Wellington IT reportedly delayed alerting customers by multiple weeks
-
OVHcloud launches European-trained open source AI modelNews The new lightweight LLM can run on standard desktop CPU, with plans to expand context window to 128,000 tokens
-
‘Organizations need to help developers grow alongside the technology’: Dejected devs lament career prospects amid AI advancesNews A new survey has found that AI is fundamentally changing the developer role and many are getting anxious
-
‘This pause is due to a significant rise in automated submissions, the vast majority of which are not valid’: Google pauses open source bug bounty scheme over AI slop submissionsNews The Google open source scheme is the latest bug bounty to fall victim to AI-generated submissions
-
Software teams should take a leaf out of manufacturers books when it comes to testing codeNews Software testers are struggling to keep up with the pace of code production. UiPath thinks it has the solution
-
Can AI fight AI? Where the security gap still exists in cybersecurity, and how MSPs can help.Industry Insights Why AI security is failing and how MSPs can close the gap
-
Microsoft may be making AI work — finally boosting its share priceMicrosoft's quarterly results featured better than expected figures on AI and cloud
-
'It delivers world-class performance at 50 percent of the cost of leading models': Microsoft unveils cut-price AI for security with latest in-house model launchNews Pairing the MAI security model with GPT-5.4 gives benchmark leading results at half the cost, according to the tech giant
-
Cisco just launched two cyber-focused small language models: Antares-350M and Antares-1B aim to supercharge codebase analysis – and they run at a “fraction of the compute expense” of popular frontier modelsNews The Antares models unveiled by Cisco aim to cut costs in codebase analysis