OpenAI Pipeline Finds 948 Linux Kernel Privilege Escalation Flaws

OpenAI's prototype cyber model pipeline identified and exploited 948 unique local privilege escalation vulnerabilities in the Linux kernel during October 2026 testing, generating candidate patches for review. The system combined automated discovery, dynamic validation, exploit generation, and human oversight on a large-scale codebase.