OpenAI's cutting-edge cybersecurity model identified approximately 1,000 Linux local privilege escalation vulnerabilities in one week.

OpenAI is deploying substantial cutting-edge computing power to freely discover and patch vulnerabilities in open-source codebases. Its upcoming advanced cybersecurity model identified approximately 1,000 local privilege escalation (LPE) vulnerabilities in the Linux kernel within a week at a cost of roughly $1 million in tokens, while also uncovering at least 10 new categories of Linux exploitation techniques; for comparison, security researchers typically discover around 20 LPEs in Linux per year. Greg Brockman added that the Linux kernel is particularly critical software for OpenAI.