OpenAI says AI pipeline generated 948 Linux exploits

OpenAI says an autonomous pipeline using an unreleased cyber model generated 948 distinct Linux kernel local privilege-escalation exploits during a run in mid-October, at a reported cost of $1,213,430 in API credits. The company said the flaws could allow a user with limited local access to gain root privileges, the highest level of control on a system. OpenAI reported that the system identified 2,308 candidate findings and dynamically reproduced 1,724 vulnerabilities within 30 hours. It generated 1,169 working local privilege-escalation exploits, then used analysis to identify 948 distinct vulnerabilities behind them. The pipeline also produced 291 proofs of concept for information leaks, OpenAI said. The company said its security researchers reviewed the exploits, proposed fixes and severity assessments before sending findings to Linux maintainers. OpenAI has submitted hundreds of reports and patches for critical flaws and plans to report remaining findings in stages with partners in its Patch the Planet initiative. The company did not provide a current total of vulnerabilities that have received upstream fixes. OpenAI said that, in a sample of 166 exploits, its agents used...