OpenAI: Cybersecurity process validates 948 Linux kernel privilege escalation vulnerabilities

OpenAI said an upcoming cybersecurity model, run through a vulnerability-research pipeline in mid-October, reproduced 1,724 Linux kernel vulnerabilities and generated working exploits and candidate patches for 948 distinct local privilege-escalation flaws. The same process produced proofs of concept for 291 information-disclosure bugs at a total API cost of about $1.21 million. OpenAI said human security researchers review the results before reporting them to maintainers, and that it has already submitted hundreds of critical bug reports and patches while working with partners on the remainder. The result matters as a test of whether AI-driven vulnerability discovery can scale across a codebase as large and widely deployed as the Linux kernel without drowning maintainers in false positives. What to watch next is how many of the submitted patches are accepted upstream and whether OpenAI extends the pipeline to other open-source projects.