OpenAI doubles bio jailbreak bounty to $50,000
OpenAI is turning its GPT-5.5 Bio Bug Bounty into an ongoing private program called the OpenAI Bio Bounty Program. The focus remains narrow: researchers will test for universal jailbreaks, meaning prompts or methods that can reliably bypass biosafety safeguards across frontier models in a predefined biological-risk challenge.
The top reward has doubled from $25,000 to $50,000 for qualifying universal jailbreaks against GPT-5.6 and GPT-5.5. OpenAI may also offer smaller rewards for partial results, depending on their value.
Testing under the original GPT-5.5 program scope will end on July 27, 2026. After that, only GPT-5.6 will remain in scope unless OpenAI announces changes. Researchers can apply through a rolling application process, and previously submitted applicants do not need to reapply.
Why it matters
- —The program targets jailbreaks that could weaken AI safeguards around biological-risk information.
- —Higher rewards may attract more expert researchers to test frontier models before broader deployment.
- —The shift to an ongoing private program suggests biosecurity testing is becoming a long-term part of AI safety work.
Key facts
- OpenAI is renaming and extending the GPT-5.5 Bio Bug Bounty as the OpenAI Bio Bounty Program.
- The top reward for a universal jailbreak is increasing from $25,000 to $50,000.
- GPT-5.5 testing under the original scope ends on July 27, 2026.
- After that date, GPT-5.6 will be the only model in scope unless OpenAI updates the program.
- Accepted researchers need ChatGPT accounts and must sign an NDA.
The full text is in the original source. Here we provide a brief summary and key facts.
