OpenAI announced updates to its Security Bug Bounty program that increase reward midpoints for medium, high and critical findings, introduce a quality-based multiplier, create a five-tier loyalty bonus system, and establish short-term priority campaign pools for newly launched features.
What changed
- The company says it has, on average, more than quadrupled the midpoint of its reward ranges for medium, high and critical vulnerabilities.\
- A critical vulnerability in core products such as ChatGPT, Codex, or the API Platform can now earn up to $30,000 as a base reward.\
- Exceptional critical findings remain eligible for a $100,000 base award and can reach up to $172,000 when applicable bonuses are applied.\
- High-quality reports that include a reliable proof of concept, clearly isolate the bug, and demonstrate exploitability and impact can receive a quality multiplier on the base bounty.\
- A new five-tier loyalty bonus program will give scaling bonuses to researchers who repeatedly partner with OpenAI.\
- Priority campaigns with dedicated bonus pools will reward early testing of newly released features.
Illustrative examples
OpenAI provided example scenarios to show how base awards combine with quality, loyalty and priority bonuses (these examples are illustrative):
-
Medium — ChatGPT (core asset): a read-only member can upload files to a Project without write permission.
- $2,000 × 1.15 Quality Bonus × 1.05 Bronze Loyalty Tier Bonus
-
High — Prism (non-core asset): IDOR resulting in reading a file uploaded to a user’s Project.
- $2,500 × 1.15 Quality Bonus × 1.10 Gold Loyalty Tier Bonus
-
Critical — ChatGPT (core asset): one-click account takeover.
- $30,000 × 1.15 Quality Bonus × 1.30 Priority Campaign Bonus × 1.15 31337 Loyalty Tier Bonus
-
Exceptional — ChatGPT (core asset): significant compromise of a security boundary (for example, the ability to view arbitrary ChatGPT users’ chat messages).
- $100,000 × 1.15 Quality × 1.30 Priority Campaign × 1.15 31337 loyalty
The company notes these scenarios and past awards are illustrative.
Retroactivity and timing
- Reports already submitted but not yet awarded will be evaluated under the new reward ranges.\
- For bounties paid in the 30 days before the announcement, OpenAI will pay the difference if the finding would have earned more under the new ranges (base award plus loyalty awards).
Engagement with the research community
OpenAI framed the changes as measures to encourage and empower independent security researchers and to reinforce a community-first security approach that complements its internal work (referred to as Defense Factory).\
The security team will attend PROMPT2PWN in Oakland on November 20–21, where researchers will be able to test targets in person and discuss collaboration opportunities.
Why it matters
Independent researchers are a key part of software security. By raising payouts and adding incentives for high-quality reports and long-term collaboration, OpenAI aims to attract more and better reports, speed up reproduction and fixes, and recognize early contributions to newly launched features. The company thanked researchers who have reported issues and said it wants to remain a partner researchers return to when they find vulnerabilities.
