OpenAI research pipeline builds 948 Linux kernel root exploits

OpenAI revealed that an autonomous system powered by an upcoming cyber-focused model generated working root exploits for 948 unique Linux kernel vulnerabilities, SecurityWeek, Help Net Security, CSO Online, and The Hacker News reported. The experiment targeted local privilege escalation flaws in the Linux kernel during a single automated run in mid-October 2026. SecurityWeek and Help Net Security reported that the target machines ran an unmodified desktop Linux operating system. [1][2][3][5]

The research run cost $1.2 million in API credits, according to SecurityWeek, Help Net Security, CSO Online, and The Hacker News. Help Net Security and CSO Online noted that the autonomous system also generated a patch for each underlying flaw. Human researchers will review the fixes before submitting them to Linux maintainers, CSO Online reported. [1][2][3][5]

SecurityWeek reported that the automated pipeline first narrowed the Linux kernel down to security-relevant areas reachable by low-privileged users. The pipeline flagged 2,308 potential bugs and confirmed 1,724 of them, eventually yielding 948 confirmed root exploits at approximately $1,280 per vulnerability. [1]

The Linux kernel is a free and open-source Unix-like kernel created by Linus Torvalds in 1991, which forms the basis for many computer operating systems. OpenAI is an artificial intelligence public benefit corporation headquartered in San Francisco, California. [11][12]