OpenAI doubles its bio bug bounty to $50,000 for a universal jailbreak

OpenAI will pay $50,000 to anyone who can break GPT-5.6’s biosecurity.

The program

On July 10, 2026, OpenAI turned its Bio Bug Bounty into an ongoing private program and doubled the top reward to $50,000, according to the company. The target is narrow and hard: find a universal jailbreak that defeats a predefined biosafety challenge against OpenAI’s frontier models.

This is not an open contest anyone can wander into. OpenAI is inviting researchers with AI red teaming, security or biosecurity experience to apply. Accepted participants submit exploits directly to the company, and OpenAI uses those attacks to harden its safeguards before wider deployment.

Why biology, and why now

The subtext of the program is the interesting part. As models get better at biology, the cost of one working jailbreak goes up. A prompt that reliably strips biosafety guardrails from a frontier model is no longer a curiosity, it is a genuine security liability.

Frontier labs have done the math on that risk, and the conclusion is pragmatic: pay serious money for attacks that arrive through the front door instead of the dark web. A $50,000 bounty is cheap compared to the alternative, which is discovering the same exploit only after it circulates in places where nobody files a responsible disclosure report.

The front-door strategy

Bug bounties are an old idea in software security, but applying them to model safety marks a shift in how AI labs think about their own products. The assumption baked into this program is that universal jailbreaks probably exist, that someone will find them, and that the only question is who gets there first and what they do with the result.

By making the program ongoing rather than a one-off challenge, OpenAI is effectively putting a standing price on its own model’s failures, and betting that the researchers capable of finding them would rather get paid and credited than go rogue.

Would you take a run at a frontier model for $50,000?