OpenAI AI Pipeline Finds and Exploits 948 Linux Root Flaws for $1.2M

· Cybersecurity · United States

OpenAI detailed a pipeline that used an upcoming cyber-focused AI model to validate and patch 948 local privilege escalation vulnerabilities in the Linux kernel, at about $1.2 million in API costs. The prototype analyzed 2,308 candidate kernel findings and dynamically reproduced them during a roughly 30-hour run, according to OpenAI. The work produced working exploits that gave root access, alongside patches. SecurityWeek reported that the model behind the effort is unreleased. The disclosure signals that AI agents can now find and weaponize real kernel bugs at a scale and speed that is difficult for human review processes to match, raising questions about how quickly defenders and open-source maintainers can absorb and remediate such output.

Why it matters：AI-driven vulnerability discovery at this scale compresses the time between bug and exploit, forcing maintainers and enterprises to rethink patch cadence and how machine-generated security findings are triaged.

Key points

  • The prototype analyzed 2,308 candidate Linux kernel findings and reproduced the flaws dynamically
  • The run took roughly 30 hours and cost about $1.2 million in API usage
  • The cyber-focused model used for the work has not been released

Sources (original reporting)

More stories from this day