OpenAI Raises Security Bug Bounty Rewards and Plans Next Internal Security Sprint
Critical bugs in core products now pay up to $30,000, exceptional findings up to $172,000 with bonuses, and security firm Cathedral joins OpenAI's next sprint.
Published
Last updated / Fact-checked on
OpenAI said on November 2 that it is "more than quadrupling the midpoint of our reward ranges for medium, high, and critical findings on average" in its Security Bug Bounty program, according to its post "Rewarding defenders with bigger bug bounties." Critical bugs in ChatGPT, Codex or the API Platform now pay up to $30,000; exceptional findings can reach $172,000 with bonuses. Pending reports get the new rates.
Photo: Pexels / Pixabay
The changes at a glance
| Reward ranges | Midpoints more than quadrupled on average (medium, high, critical) |
| Critical, core products | Up to $30,000 |
| Exceptional findings | Up to $100,000 base; up to $172,000 with bonuses |
| New bonuses | Quality multiplier, 5-tier loyalty, priority campaigns |
| Retroactive | Pending reports; top-ups for awards in prior 30 days |
| Security sprint partner | Cathedral |
| Defender access | $1B subsidized access over six months |
| Sources | OpenAI (two posts), Google |
What changed
OpenAI said it is raising rewards to encourage independent researchers to pursue vulnerabilities its own testing may have missed. The $172,000 ceiling applies only to exceptional findings, not to ordinary critical bugs. The bounty work complements OpenAI's "Defense Factory," where its engineers use its models to find, reproduce, and fix weaknesses, the company said.
New bonus structures
High-quality reports that include a reliable proof of concept are eligible for a quality multiplier on the base bounty. The company is also launching a five-tier loyalty bonus program. Priority campaigns will include dedicated bonus pools and multipliers for newly launched features.
OpenAI provided illustrative examples of the updated calculations. A medium ChatGPT bug moves from an illustrative historical award of $750 to $2,415, calculated as $2,000 × 1.15 quality × 1.05 Bronze loyalty. A critical one-click ChatGPT account takeover moves from $6,500 to $51,577, calculated as $30,000 × 1.15 quality × 1.30 priority campaign × 1.15 "31337" loyalty tier. An exceptional security-boundary compromise, such as the ability to view arbitrary users' chats, moves from $100,000 to $171,925.
Retroactive pay and in-person testing
Reports already submitted but not yet awarded will receive the new ranges, OpenAI said. For bounties awarded in the 30 days before the announcement, the company will pay the difference if the finding would have earned more under the new ranges, covering the base award plus loyalty awards.
OpenAI's security team will be at PROMPT2PWN in Oakland on Nov. 20-21, where researchers can test targets in person.
The Cathedral security sprint
In a second post, "Preparing our next security sprint to harden OpenAI," the company said it is planning its next internal security sprint and partnering with Cathedral, a security firm (some coverage calls it Cathedral Systems). Cathedral's researchers will use OpenAI's models alongside OpenAI's teams to challenge its defenses, OpenAI said.
Gavin Kliger, CEO and co-founder of Cathedral, stated: "OpenAI is setting an important example by partnering with outside security firms and using its own models to continuously test and strengthen its defenses. Every company should be thinking about how to put these capabilities to work."
In its last sprint, OpenAI dedicated 25% of its product engineers to finding and fixing weaknesses, using Codex to investigate findings, reproduce vulnerabilities in isolated environments, and prepare patches for review, the company said. OpenAI is making its most capable models available to Daybreak customers and has committed $1 billion in subsidized access over six months for frontline defenders protecting essential services and digital infrastructure.
The announcements come days after Google announced Gemini 4 Carbon on Oct. 28, 2026, which it calls its strongest model with state-of-the-art cybersecurity capabilities, initially available to a subset of cyber defenders through its Fairwind Program, according to Google's blog.
What to watch
- PROMPT2PWN. What researchers find at the Nov. 20–21 Oakland event.
- Sprint results. Whether OpenAI publishes findings from the Cathedral sprint, as it did lessons from the last one.
- Priority campaigns. Which newly launched features get the first dedicated bonus pools.
- Rival programs. Whether other AI labs raise their own bounties.
More coverage is on our Artificial Intelligence hub.
How we reported this
Bounty figures, examples, retroactive terms and sprint details come from OpenAI's two Nov. 2 posts; the examples are OpenAI's illustrations, not actual payouts. The Gemini 4 Carbon details come from Google's announcement. OpenAI's post on X could not be retrieved and was not used.
Published and fact-checked November 2, 2026.
Sources
- OpenAI, "Rewarding defenders with bigger bug bounties," Nov. 2, 2026: https://openai.com/index/rewarding-defenders-with-bigger-bug-bounties/
- OpenAI, "Preparing our next security sprint to harden OpenAI," Nov. 2, 2026: https://openai.com/index/preparing-our-next-security-sprint-to-harden-openai/
- Google, "Introducing Gemini 4 Carbon," Oct. 28, 2026: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon/
DailyMech used AI tools to help research and draft this article; it was fact-checked by our editors.