AI BriefingGPT-5.5 Bio Bug Bounty
Key point
OpenAI has launched a bug bounty to find jailbreaks that bypass GPT-5.5's bio safety measures.
Details
OpenAI is running a Bio Bug Bounty program targeting GPT-5.5. The goal is to find a universal jailbreak that passes all 5 bio safety questions.
The scope is limited to GPT-5.5 on Codex Desktop only. Participants must find a way to get all five questions answered with a single prompt in a clean, new chat, without triggering moderation.
The reward is $25,000 for the first genuine universal jailbreak that passes all 5 questions. Smaller prizes may be awarded at OpenAI's discretion for partial successes.
The schedule is as follows:
- Applications: April 23 - June 22, 2026
- Testing: April 28 - July 27, 2026
Participation combines applications and invitations. OpenAI will send invitations to trusted bio red-teamers, and also review new applications to onboard selected participants. Approved participants and collaborators must have a ChatGPT account and sign an NDA, which covers prompts, completions, findings, and communications throughout.
Separately, OpenAI is also running a Safety Bug Bounty and a Security Bug Bounty.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Check Point Reveals 'PuzzleMask' Attack Technique That Bypasses LLM Security Gatekeepers Using Plain Text