OpenAI's autonomous kernel vulnerability pipeline

OpenAI is submitting patches to Linux kernel maintainers through its 'Patch the Planet' initiative after an autonomous pipeline found, validated, and generated fixes for 948 local privilege-escalation vulnerabilities in the kernel. Within 10 days of the scan running, over 100 Linux LPEs had been patched or had staging patches from other contributors, with around 300 more publicly discussed. The pipeline used AI scout agents to narrow the kernel's 40 million-plus lines of code to approximately 2.8 million lines, about 7 percent, considered most security-relevant; it then flagged 2,308 suspected bugs, confirmed 1,724 on test machines, and produced working exploits and patches for 948. The pipeline was also tested against a hardened version of Google's kernelCTF, producing 21 LPE exploits over 112 hours at roughly $33,000 in API credits. Human reviewers checked results before patches were submitted. Total API cost for the main run was $1,213,430, with $775,929 spent on exploit generation; because OpenAI used its own model, that figure may not reflect actual cash outlay. The pipeline used an upcoming, unreleased cyber-focused model.