OpenAI announced on 2 November 2026 that it plans a security sprint with Cathedral Systems and other outside researchers, alongside bug bounty changes that can bring an exceptional critical payout to about $172,000, Unite.AI reported. Cathedral researchers are expected to work with OpenAI teams to find vulnerabilities and test fixes using the company’s models.
Key points
- The new sprint is planned with Cathedral Systems and other external researchers.
- Critical findings in core products can earn up to $30,000. Exceptional findings can receive a $100,000 base award before bonuses.
- OpenAI says checks after an earlier sprint found that merged patches had not always reached deployed systems.
Cathedral joins OpenAI’s planned sprint
Gavin Kliger, Cathedral Systems’ CEO and co-founder, said the joint effort would draw on his researchers’ work as well as OpenAI’s models and security expertise. The work is intended to find weaknesses and check repairs. Cathedral’s involvement adds external testing to an internal process OpenAI has already used to investigate reports and prepare patches.
The announcement concerns a forthcoming sprint rather than findings from one. Cyber Security News reported that the announcement gave no start date, target list or results for the planned work. OpenAI’s teams and the outside researchers are expected to challenge existing protections together.
OpenAI used Codex during its earlier sprint to investigate suspected flaws, reproduce them in isolated environments and prepare patches for engineers to review. Its Defense Factory process also assigns possible vulnerabilities to the teams responsible for the affected systems. That workflow gives the Cathedral researchers an existing route for getting a finding checked and a proposed repair reviewed.
Critical bounties reach $30,000 before exceptional awards
OpenAI’s revised bug bounty pays up to $30,000 for a critical vulnerability in a core product such as ChatGPT, Codex or the API Platform. Exceptional critical findings remain eligible for a base award of up to $100,000. Bonuses for report quality, researcher loyalty and priority campaigns can bring a payout to about $172,000, Cyber Security News reported.
OpenAI says the changes more than quadrupled the midpoint of its reward ranges for medium, high and critical findings on average, Unite.AI reported. A report can qualify for a quality bonus when its proof of concept isolates the flaw, establishes when it can be exploited and demonstrates the impact. The larger advertised payout therefore applies to an exceptional finding with additional bonuses, rather than to every critical report.
A new five-tier loyalty programme raises bonuses for researchers who work with OpenAI repeatedly. Priority campaigns will have separate bonus pools for testing newly launched features. Reports awaiting an award move to the new ranges, while researchers awarded bounties in the 30 days before the announcement will receive the difference where the revised base and loyalty awards would have paid more.
OpenAI has previously paid a $5,000 bounty after a ChatGPT share-link flaw exposed private chats. The revised programme gives independent researchers a higher potential payment for serious findings, with the amount dependent on the finding and the applicable bonuses.
OpenAI’s earlier sprint tested deployed fixes
OpenAI assigned 25% of its product engineers to the earlier security sprint. Its account of the effort put more than 250 people across more than 100 service areas. The company says it closed 53 urgent or high-priority issues on the first day and classified 37% of findings as duplicates, Unite.AI reported.
OpenAI says it reproduced 19.5% of findings at runtime and recorded a 0.81% false-positive rate after dynamic validation. Codex prepared the remediation work, which then went through review. Isolated test environments let engineers check suspected flaws without running those experiments against live systems.
Follow-up checks found that some patches merged into the code had not reached every deployed system, according to OpenAI’s account reported by Cyber Security News. The company’s workflow consequently includes human review and checks of the deployed fix, beyond accepting a patch for the codebase.
_-2.jpg){kind=link}