OpenAI has developed an AI-powered cybersecurity system that identified 948 distinct Linux kernel privilege escalation vulnerabilities and generated working exploits.
OpenAI has developed an AI-powered cybersecurity system that identified 948 distinct Linux kernel privilege escalation vulnerabilities and generated working exploits.
The system analyzed 2,308 potential security issues, reproduced 1,724 vulnerabilities, and automatically generated candidate patches. The experiment cost approximately $1.2 million in API usage, with human researchers reviewing the findings and proposed fixes.