The pipeline analyzed 2,308 candidate findings, reproduced 1,724 vulnerabilities, and produced 948 local privilege escalation exploits, which are attacks that let an unprivileged local user gain root access.
Human researchers reviewed the results, and OpenAI says it has already sent hundreds of reports and patches for critical issues to the Linux community.
It is working with Patch the Planet partners to release the rest gradually.
OpenAI says the hard part was no longer finding flaws but deciding which could be reproduced, what security boundary they crossed, and which deserved maintainers' attention first.
The company also compared costs with bug bounty payouts.
Applying historical Pwn2Own and TyphoonPWN rewards to the 948 exploits gives a hypothetical value of $18.9 million to $66.3 million, or roughly 16 to 55 times the pipeline's cost.
In a run against Google's hardened kernelCTF environment, it produced 21 local privilege escalation exploits for about $32,936 in API credits.
OpenAI says successful exploits averaged $359 each to develop and describes its reported costs as an upper bound.
The system works in five stages: discovery, dynamic validation in isolated virtual machines, exploitability assessment, patch generation, and human review. Scout agents narrowed the audit to about 7% of the kernel, and about three quarters of candidates were reproduced, with 3.1% reflecting incorrect model analysis. Maintainers moved quickly. After a 10-day scan of the latest kernel, 61 privilege escalation flaws were already patched upstream, 43 had staging patches, and 299 were under public discussion. OpenAI also reports that 20 of 200 early "successful" exploits had gained root through a different, unassigned 0-day flaw, which led it to add independent checks. It says agents used 10 exploitation techniques that were new to outside kernel experts, and that it is withholding the technical details for safety. It plans to extend the work to more critical open-source projects through Patch the Planet.
