OpenAI says its new cyber-focused AI model found thousands of Linux kernel vulnerabilities

OpenAI’s new GPT-6 Cyber LLM found hundreds of Linux kernel vulnerabilities, the company claimed, touting the model’s improved bug-hunting capabilities.

Unveiled on 5 November, the company said the new model marks a “major leap in security research” and is more capable of finding and fixing vulnerabilities than its predecessor.

“GPT-6 Cyber delivers major gains on key cybersecurity benchmarks over GPT-5.6 Cyber, including a 51% relative improvement on ExploitGym and a step function improvement when tested on staying within scope,” OpenAI said in a blog post.

OpenAI noted that the model is “already helping to secure real-world software”, and has been used to uncover “thousands” of exploits spanning Linux and Firefox.

With the latter, GPT-6 Cyber uncovered 81 high-severity vulnerabilities, also helping to create patches. These have since been patched by the browser’s creator, Mozilla.

GPT-6 Cyber flags Linux kernel flaws

OpenAI specifically highlighted the volume of Linux kernel vulnerabilities flagged by the new cybersecurity-focused model, racking up 2,308 potential flaws and reproducing some 1,726 vulnerabilities.

Notably, the company revealed the model helped develop local privilege escalation exploits for 948 vulnerabilities during testing. These could allow an unprivileged user to gain root access.

The announcement from OpenAI comes as Linus Torvalds, the creator of Linux kernel, voiced cautious optimism in terms of the impact of AI on vulnerability disclosures. In a weekly kernel update published on 8 November, Torvalds noted that a recent release saw “various AI fixes”.

Torvalds added: “Fixes are now coming in faster than the new reports, and that sounds lovely if true”.

With the launch of powerful new cyber-focused models by OpenAI and Anthropic in recent months, organizations globally have battled a wave of previously unknown vulnerabilities.

Figures published by Google Threat Intelligence Group (GTIG) in October showed the number of vulnerability disclosures each month more than doubled since the start of the year.

GTIG also warned the number of vulnerabilities being exploited showed a similar trajectory.

In April this year, researchers at Forescout told ITPro that enterprises should brace for vulnerability explosion, warning that many could struggle to keep pace with the volume of vulnerability disclosures.

In a blog post detailing the Linux vulnerability research, OpenAI said that the rate of vulnerability discovery by frontier models means automated patch generation is now “essential to secure software at the pace required and to limit the exploitation window”.

FOLLOW US ON SOCIAL MEDIA

Follow ITPro on Google News and add us as a preferred source to keep tabs on all our latest news, analysis, views, and reviews.

You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.