OpenAI's Bio Bug Bounty Reward Doubles to $50,000!

Hi, I'm Shii-chan! Today I want to share an AI safety story, and it's a bug bounty program that gets me a little excited.

OpenAI News openai.com

What was announced?

OpenAI News announced that its "GPT-5.5 Bio Bug Bounty" has evolved into an ongoing, private program called the "OpenAI Bio Bounty Program." The goal stays the same as before: finding a "universal jailbreak" that can defeat the safeguards around biological risk (biorisk).

The story so far

The original "GPT-5.5 Bio Bug Bounty" accepted applications on a rolling basis starting April 23 of this year, and closed on June 22. It targeted GPT-5.5 in Codex Desktop only, with a testing window that ran from April 28 through July 27.

The challenge was to find a single universal jailbreak prompt that could get answers to all five bio safety questions from a clean chat, without triggering moderation. Participation was limited to people invited from a vetted list of trusted bio red-teamers, or new applicants who passed review. All prompts, completions, findings, and communications were covered by an NDA (non-disclosure agreement).

The reward was $25,000 for the first person to find a genuine universal jailbreak that cleared all five questions, with smaller awards possible for partial successes at OpenAI's discretion.

What changes

The biggest change in this announcement is a major reward increase. The bounty for finding a universal jailbreak doubles from $25,000 to $50,000. GPT-5.6 becomes the priority target model, while GPT-5.5 stays in scope until July 27, 2026. OpenAI says scope changes for models will continue to be communicated on a rolling basis going forward.

Moving from a one-time campaign to an always-on private program means biosecurity experts can keep probing for gaps in the safeguards continuously.

Dive Deep

The announcement also covers how to apply. Interested participants need to submit a brief application form through Google Forms, including their name, affiliation, and experience. Selected applicants will need a ChatGPT account and will be asked to sign an NDA. People who already applied to the previous program don't need to reapply.

Wrap-up

• The "GPT-5.5 Bio Bug Bounty" has evolved into an ongoing private program called the "OpenAI Bio Bounty Program"
• The reward for finding a universal jailbreak doubles from $25,000 to $50,000
• GPT-5.6 is now the priority target model, with GPT-5.5 in scope until July 27, 2026
• Applications go through Google Forms, and selected participants need a ChatGPT account plus an NDA

This is a must-know update for researchers with expertise in biosecurity or AI red teaming!