OpenAI: Cybersecurity Workflows Verify 948 Linux Kernel Privilege Escalation Vulnerabilities

on November 5, OpenAI disclosed that it uses an upcoming cybersecurity model to build a vulnerability research workflow. In a run during mid-October, it reproduced 1,724 Linux kernel vulnerabilities and generated working exploit code and candidate patches for 948 distinct local privilege escalation vulnerabilities.

The workflow also generated proofs-of-concept for 291 information disclosure vulnerabilities, with a total API cost of approximately $1.21 million. OpenAI stated that security researchers will review the results and report vulnerabilities to maintainers. The company has submitted hundreds of critical vulnerability reports and patches, and is working with partners to advance the remaining patches.