OpenAI says its AI vulnerability-hunting pipeline validated 948 Linux kernel privilege-escalation flaws in a single run

November 5: OpenAI posted on X that its security researchers built a cybersecurity pipeline that can autonomously discover, exploit, and patch vulnerabilities. In a single run paired with an upcoming cybersecurity model, the pipeline reproduced 1,724 vulnerabilities in total, generated proofs of concept for 291 information-disclosure vulnerabilities, and produced working exploit code and candidate patches for 948 distinct Linux kernel local privilege-escalation vulnerabilities, at a total API cost of about $1.21 million, or roughly $1,279 per privilege-escalation vulnerability. OpenAI said the agent also discovered and used at least 10 previously unknown Linux exploitation techniques, and is currently working with Linux maintainers and Patch the Planet partners to advance vulnerability reports and patches.