OpenAI offers $25,000 to break GPT-5.5's safety features
25 Apr 2026

OpenAI has launched a "Bio Bug Bounty" program, offering a reward of $25,000 for security researchers who can bypass the safety measures of its new AI model, GPT-5.5.

The initiative is part of OpenAI's efforts to ensure AI safety and is one of the first cases where a major tech company is using external adversarial testing for this purpose.

'Universal jailbreak' prompt needed for biosafety challenge
Program details

The Bio Bug Bounty program, which opened for applications on April 23, challenges participants to find a universal "jailbreak" prompt.

This prompt should be able to get the model to answer all five questions in a biosafety challenge without triggering any moderation response.

The task has to be done from a clean chat session, with no prior conversation or context that could influence the model's responses.

Partial successes may also be rewarded
Reward structure

The first researcher to achieve a complete universal jailbreak across all five questions will be rewarded with $25,000.

OpenAI may also reward partial successes at its discretion, although the amounts have not been specified.

The company has said that applications for the program will close on June 22, 2026, and will be reviewed on a rolling basis.

Access limited to vetted group of trusted biosecurity red teamers
Participant criteria

OpenAI has made it clear that access to the Bio Bug Bounty program isn't open to all.

The company will invite a vetted group of trusted biosecurity red teamers and also consider applications from researchers with relevant experience in AI red teaming, security, or biosecurity.

All findings, prompts and communications will be covered by a non-disclosure agreement (NDA), meaning participants can't publicly disclose their results.