OpenAI pipeline produces 948 working Linux kernel privilege-escalation exploits
On 4 November 2026 OpenAI published results from an autonomous vulnerability-research pipeline built on an unreleased cyber model. In a single run in mid-October 2026 against the Linux kernel master branch, the pipeline produced 2,308 candidate findings, dynamically reproduced 1,724 vulnerabilities, and generated 291 information-leak proofs of concept and 948 working local privilege-escalation exploits, each paired with a candidate patch. The run cost $1,213,430 in API credits, roughly $1,279 per exploit. On an adapted version of Google's hardened kernelCTF environment the same pipeline obtained 25 information-leak proofs of concept and 21 LPE exploits.
Why It Mattered
This is the clearest public measurement to date of what machine-speed vulnerability research costs. OpenAI contrasted its figure with historical Pwn2Own and TyphoonPWN bounties of $20,000 to $70,000 for a Linux kernel LPE, implying the 948 exploits would carry $18.9M to $66.3M in bounty value, 16 to 55 times the pipeline's cost. The relevant quantity for the history of AI is not the exploit count but the unit price: finding root on the most widely deployed kernel in the world moved from a scarce specialist achievement to a budget line. The result also sets a defensive baseline, since every exploit arrived with a root-cause analysis and a patch, and the company routed findings through human security reviewers to kernel maintainers before publication. That creates an asymmetry problem that regulators and maintainers will be arguing about for years: a volunteer-maintained codebase cannot absorb patches at the rate an automated pipeline can generate findings, and the same capability is available to anyone who can rent comparable compute. The disclosure lands in a year when several labs reported their models being used in real intrusions, and it converts an argument about hypothetical offensive capability into an auditable number. For 2030 retrospectives on when AI changed the economics of software security, this writeup is the citable artifact.
Who Built It
OpenAI
Applications
- Cybersecurity
- Vulnerability Research
- Software Maintenance