Nov 4, 2026
OpenAI reports that a prototype using an unreleased cyber model reproduced 1,724 Linux kernel vulnerabilities and produced 948 distinct local privilege escalation exploits in a controlled run.
The mid-October experiment combined discovery, validation, exploitability testing and candidate patch generation. Human researchers reviewed findings and prioritized disclosures. The 948 candidate patches were prepared for review; upstream acceptance and deployment are separate milestones.