OpenAI says AI pipeline found 948 Linux kernel flaws and generated 1,169 LPE exploits
Summary
OpenAI says an AI pipeline found 948 Linux kernel flaws and generated 1,169 working local privilege escalation exploits in a scan costing $1.2 million; it has submitted hundreds of reports and patches.
Key points
- OpenAI says the October scan flagged 2,308 potential bugs and confirmed 1,724; it reports 948 distinct patches for bugs behind the exploits.
- The pipeline produced 1,169 working local privilege escalation exploits and 291 information-leak proofs of concept, tested in virtual machines with common protections enabled.
- OpenAI says the pipeline uncovered at least 10 novel Linux exploitation techniques, but is withholding full technical details.
- A scan against an adapted, hardened kernelCTF target produced 21 LPE exploits over 112 hours at a cost of about $33,000.
- OpenAI says it has submitted hundreds of reports and patches and will send remaining fixes in stages; it also recommends keeping systems patched and using sandboxing and isolation.
- Separately, OpenAI raised Security Bug Bounty payouts, with rewards for critical flaws in core products reaching $30,000 and exceptional payouts potentially reaching $172,000.
Article Details
- Event Type
- OpenAI reported using an autonomous AI-agent pipeline to discover and test Linux kernel vulnerabilities, and separately announced increased payouts in its Security Bug Bounty program.
- Impact
- OpenAI reported that the pipeline produced 1,169 working local privilege escalation exploits and 291 information-leak proofs of concept; 948 separate patches addressed the bugs behind the exploits. OpenAI said it had submitted hundreds of reports and patches for critical flaws, with remaining fixes planned in stages. It also warned that an unpatched system accessible to a low-privileged user could be at risk of privilege escalation. The expanded bounty program offers up to $30,000 for qualifying critical flaws, with bonuses potentially raising a single payout to $172,000.
Vendors
Products
API PlatformA researcher who reports a critical flaw in ChatGPT, Codex, the API Platform or another core product can now earn as much as $30,000.ChatGPTA researcher who reports a critical flaw in ChatGPT, Codex, the API Platform or another core product can now earn as much as $30,000.CodexA researcher who reports a critical flaw in ChatGPT, Codex, the API Platform or another core product can now earn as much as $30,000.GPT-5.6 SolSpread across the 948 exploits, the $1,213,430 bill comes to roughly $1,279 apiece. An earlier scan in late July, powered by GPT-5.6 Sol, yielded 166 unique LPEs at a cost of $513,690.Linux kernelOpenAI says an autonomous pipeline powered by an unreleased cyber model built working root exploits for 948 unique Linux kernel vulnerabilities during a single run that cost $1.2 million in API credits.