OpenAI has announced a new challenge for researchers who think they can bypass the biosafety protections built into the GPT-5.6 AI model. The company has updated its its Bio Bounty Programme, and is now offering rewards of up to $50,000 (nearly Rs 47 lakhs) to those who can find security weaknesses in the AI system before it could be exploited in the real world. This initiative expands on the previous GPT 5.5 programme, raising the maximum reward and will become the company’s primary biosafety testing after July 27, 2026.
What Is OpenAI’s Bio Bug Bounty Programme?
The Bio Bug Bounty programme can identify “universal jailbreaks” that can consistently defeat GPT 5.6’s biology-related safety guards. Rather than looking for prompts that work only once or in a specific conversation, the company wants researchers to discover techniques that reliably bypass its protective systems across different interactions.
OpenAI has doubled the top reward for successful submissions. Researchers who discover a qualifying universal jailbreak can get up to $50,000 (nearly Rs 47 lakhs), compared to the earlier maximum of $25,000. Moreover, the company has asserted that it will offer small payouts for partial findings. The final reward will be based on the quality, originality and potential impact of each submission.
How To Apply
Those interested in joining the programme can visit OpenAI’s Bio Bug Bounty webpage and click on the Apply Now button. Applicants will be required to offer basic information, including their name, affiliation and relevant experience. They must also have a ChatGPT account and sign a non-disclosure agreement (NDA) before getting access to the testing platform. OpenAI also clarified that researchers who have already applied to the GPT 5.5 Bio Bug Bounty Programme do not need to submit a fresh application for GPT 5.6.
The expanded bounty programme also reflects a growing industry trend of using independent security researchers to improve the reliability and safety of advanced AI models before they reach more users.
