What happened
OpenAI (July 9, 2026) converted its time-boxed GPT-5.5 Bio Bug Bounty into an ongoing private program (OpenAI Bio Bounty Program) testing for universal jailbreaks against biosafety safeguards on frontier models starting with GPT-5.6, raising the reward from $25,000 to $50,000.
Why it matters
Formalizes continuous red-teaming of a high-consequence risk category (bio-uplift) as models scale, signaling frontier labs are institutionalizing adversarial safety testing rather than one-off bounties.
Applicability
AI safety/red-teaming teams at frontier labs and enterprises deploying frontier models in regulated/high-risk domains should track scope and reward structure as a benchmark for their own red-teaming programs.