OpenAI doubles bio jailbreak bounty to $50K

OpenAI has doubled the top reward in its Bio Bounty Program to $50,000 for researchers who can craft a reusable jailbreak against its biological safety safeguards. The program initially targets GPT-5.6, with GPT-5.5 remaining in scope until July 27, 2026. This initiative aims to stress-test biosecurity controls before enterprises deploy advanced models in sensitive research environments. Participation is restricted: applicants must submit their credentials, sign an NDA, and use OpenAI’s bounty platform. Accepted researchers gain access to a predefined biosafety challenge where a single prompt must bypass safeguards to answer five biosecurity questions without triggering moderation.

OpenAI’s move follows earlier vulnerabilities, including a UK AI Security Institute test where a universal jailbreak produced policy-violating responses in six hours. While that test focused on cybersecurity, OpenAI’s program specifically targets biological safety gaps. The company’s system card for GPT-5.5 noted that sustained expert jailbreaking revealed model-level biological safety failures during predeployment testing, though safeguards blocked verified high-severity attempts. For enterprises, this underscores the need to treat vendor safeguards as one layer of protection—complemented by least-privilege controls, audit trails, and human review—especially as AI systems gain access to critical business tools.