OpenAI Cyber Model Finds 948 Linux Kernel Privilege Escalation Flaws

OpenAI says an automated security research prototype found 948 unique Linux kernel vulnerabilities that could let an unprivileged local user gain root access, as detailed in a company blog post. The system identified 2,308 candidate findings, reproduced 1,724 vulnerabilities and produced 1,169 working privilege escalation exploits that corresponded to 948 distinct patches.

The pipeline, powered by an upcoming OpenAI cyber model, expands its Patch the Planet work from June, when GPT-5.5 Cyber produced 24 such exploits. It used agents to select 5,034 audit targets covering about 2.8 million lines of code, or roughly 7% of the Linux kernel. Other agents scanned the targets, reproduced suspected flaws in isolated virtual machines, assessed exploitability and generated candidate patches. Human researchers then reviewed the exploits, patches and proposed severity before reporting findings to maintainers.

The October 2026 scan cost $1,213,430 in API credits. The exploitation stage accounted for $775,929, while successful exploits cost an average of $359 to develop. Tests used standard Linux protections including address randomization and controls that restrict kernel access to user memory.

The scan examined the latest Linux kernel code available in mid October. By the time automated scanning and manual analysis finished 10 days later, 61 of the privilege escalation flaws had been patched, 43 had staging patches and 299 were being reported or discussed publicly. OpenAI has sent hundreds of reports and candidate patches to Linux maintainers, starting with critical vulnerabilities.