claudeenmodel: claude-haiku-4-5
GPT-5.5 Bio Bug Bounty Program Launched
Key Points
- Universal jailbreak challenge with $25,000 reward
- GPT-5.5 biosafety testing program
- Rolling applications through June 22, 2026
Summary
OpenAI has launched a Bio Bug Bounty program for GPT-5.5 to identify universal jailbreaks that could defeat biosafety protections. Researchers with AI red teaming, security, or biosecurity expertise are invited to participate in this security testing initiative.
Key Points
- Model & Scope: GPT-5.5 in Codex Desktop only
- Challenge: Find a single universal jailbreaking prompt that successfully answers all five bio safety questions without triggering moderation
- Reward: $25,000 for the first true universal jailbreak; partial wins may receive smaller awards
- Timeline: Applications open April 23, 2026 and close June 22, 2026; testing runs April 28 – July 27, 2026
- Eligibility: Vetted bio red-teamers and reviewed applicants with existing ChatGPT accounts
- Requirements: Participants must sign an NDA covering all prompts, completions, and findings
- Application: Submit name, affiliation, and relevant experience via the application portal