OpenAI Says Its Upcoming Cyber Model Found 948 Privilege-Escalation Bugs in Linux

By Theo Lin, AI reporter · AI desk

OpenAI said it used an upcoming cyber model to autonomously scan the latest Linux kernel, as of “mid-October 2026,” and found 948 local privilege-escalation vulnerabilities, The Verge reported on Nov. 6. Such bugs can let a user or program that already has access to a machine gain higher-level control.

Unconfirmed: the figure comes from OpenAI and has not been independently verified. The Boston Report has not seen a list of the bugs or a response from Linux kernel maintainers.

Related: OpenAI Says GPT-6 Cyber Lifts ExploitGym Score to 48.5%.

What OpenAI said

OpenAI’s system also generated patches for the bugs, and a human researcher reviewed them, The Verge reported, citing the company. The Verge’s brief report did not say how many of the patches have been submitted to or accepted by kernel developers, or how severe the bugs are.

Why it matters

The Linux kernel runs most of the world’s servers and Android phones. AI companies have been racing to show that their models can find and fix security flaws at scale, as The Boston Report has reported. The same skills raise concerns that attackers could use similar tools, a risk highlighted this week by reports on the exposed Artex hacking tool.

What to watch

Independent confirmation would come from kernel maintainers, CVE records or fixes landing in kernel releases. The Boston Report will update this story if OpenAI publishes technical details. More coverage is on the AI news page.

Sources

• The Verge: OpenAI says it found 948 local privilege escalation vulnerabilities in the Linux kernel

About the reporter: Theo Lin is an AI reporter covering AI, technology and science for The Boston Report. This article was produced with AI under our editorial policy. Read our AI and media disclosure.