OpenAI doubles reward for finding biological jailbreaks to $50,000

On July 9, 2026, OpenAI transitioned the GPT-5.5 Bio Bug Bounty program to a permanent closed format and doubled the maximum reward for universal biological safety bypass techniques — from $25,000 to $50,000.

How the New Program Works

OpenAI renamed the initiative to OpenAI Bio Bounty Program and made it permanent. In the program, researchers attempt to find universal jailbreaks — methods to bypass a predetermined set of biological safety constraints that work not in a single specific query but systematically against the company's models.

OpenAI's leading models starting with GPT-5.6 will undergo testing. The program's goal is to identify weaknesses in safety measures before malicious actors can exploit them to obtain dangerous biological information or instructions.

• On July 9, 2026, OpenAI announced the launch of a permanent closed OpenAI Bio Bounty Program.
• The reward for universal safety bypass on GPT-5.5 and GPT-5.6 increased from $25,000 to $50,000.
• Testing of GPT-5.5 model in the previous program will conclude on July 27, 2026.
• After July 27, 2026, GPT-5.6 will be the model under review.
• For partial results, OpenAI may award smaller payments at its discretion.

The program is not a public competition where one can directly submit a discovered bypass. Those interested must submit a brief application with their name, affiliation, and description of experience. Selected participants will be given access to a special testing platform. Researchers who previously applied to GPT-5.5 Bio Bounty will not need to resubmit an application.

What Will Researchers Search For?

Researchers will test whether a single reproducible method can remove the specified biological risk constraints. OpenAI calls such findings universal jailbreaks: they refer not to a single failed prompt formulation but to a vulnerability capable of consistently circumventing the model's protective mechanisms.

The company has not disclosed the technical specifications of its biological safety test suite. This is expected: publishing exact criteria and examples could facilitate the development of bypasses outside the controlled program. Participants will work in a closed environment, and participation requires an active ChatGPT account and signing a non-disclosure agreement.

The increased payout makes searching for such problems significantly more motivating for model safety assessment specialists. Simultaneously, the closed format gives OpenAI the ability to verify participant qualifications, control the testing environment, and process results before potential vulnerability details become public.

Why the Focus on GPT-5.6

After July 27, 2026, the program will stop accepting results for GPT-5.5 and shift to GPT-5.6. OpenAI explicitly stated that this model will be the starting point for continued permanent testing, and the company will notify researchers separately about future changes to the testing scope.

This transition shows that biosafety testing will accompany the release of new advanced models, rather than remaining a one-off campaign around GPT-5.5. For OpenAI, this is a way to receive external validation of its constraints: the internal team defines the protective framework, independent specialists search for repeatable methods to bypass it.

The Bio Bounty program exists separately from other OpenAI vulnerability discovery initiatives. The company also directs researchers to a general security vulnerability discovery program and a Security Bug Bounty program. The new format focuses specifically on risks arising from the application of advanced AI models in the biological domain.

What This Means

OpenAI is raising the price for detecting the most dangerous bypasses and transitioning biosafety testing to a continuous process. For the industry, this is a signal: assessment of risks in advanced models increasingly becomes permanent external review rather than a phase that ends with the product release.