OpenAI has launched the GPT-5.5 Bio Bug Bounty, inviting AI red teams, security researchers, and biosafety experts to try and crack GPT-5.5's biosecurity guardrails. The challenge: start a clean session on Codex Desktop, then use a single universal jailbreak prompt to answer all five biosecurity questions without triggering the content moderation system. The first person to pull it off gets $25,000. Partial breakthroughs earn smaller prizes.
Applicants must submit their name, affiliation, and relevant experience. If approved, they sign an NDA. All prompts, model outputs, and findings are covered by that NDA. Applications close June 22, with the testing window running from April 28 to July 27.