OpenAI Doubles Biosecurity Bug Bounty to $50K Amid AI Safety Concerns

• •OpenAI increased its Bio Bug Bounty reward from $25,000 to $50,000 for universal jailbreaks.
• •The program now operates as a permanent initiative, focusing on ongoing evaluations of AI models.
• •Participants must apply through a selective process and sign NDAs, limiting public visibility into vulnerabilities.

OpenAI has revamped its Bio Bug Bounty program, increasing the maximum reward for discovering universal jailbreaks from $25,000 to $50,000. This initiative aims to enhance the security of its AI models, particularly in the biological sciences. The program now operates on an ongoing basis, focusing on vulnerabilities in both GPT-5.6 and GPT-5.5 models until the latter's testing window closes on July 27, 2026. Participants must apply through a selective process and sign a non-disclosure agreement, indicating the sensitive nature of the findings. The program reflects a shift in OpenAI's approach to AI safety, emphasizing collaboration with external experts to identify potential risks. This move follows previous efforts in targeted bio-risk red teaming and highlights the growing importance of biosecurity in AI development. The initiative is part of OpenAI's broader commitment to ensuring the safe deployment of advanced AI technologies.