OpenAI turns GPT-5.5 Bio Bug Bounty into a permanent program | Keryc
OpenAI is reshaping its biosafety initiative: the GPT-5.5 Bio Bug Bounty becomes a private, ongoing program called the OpenAI Bio Bounty Program. The goal is straightforward: strengthen safeguards against dangerous uses of advanced models in biology, starting with GPT-5.6 while keeping GPT-5.5 in scope until July 27, 2026.
What changes with the OpenAI Bio Bounty Program
The program stops being temporary and becomes a private, continuous program aimed at selected researchers.
The focus remains on finding "universal jailbreaks", meaning generic ways to bypass safety restrictions that defeat OpenAI's predefined biosafety challenge for its frontier models.
Rewards for a "universal jailbreak" increase from USD 25,000 to USD 50,000 for both GPT-5.6 and GPT-5.5.
Smaller prizes may be awarded for partial findings, at OpenAI's discretion.
The testing window for GPT-5.5 ends on July 27, 2026; after that date, only GPT-5.6 will be in scope.
The application process is rolling. Accepted researchers will be integrated into the bug bounty platform and must have ChatGPT accounts and sign an NDA.
If you already applied to the GPT-5.5 Bio Bounty, you don’t need to reapply.
What does "universal jailbreak" mean and why does it matter?
You might be wondering what a "universal jailbreak" actually is. In short, it’s a technique that lets someone bypass a model’s safety barriers in a broad, repeatable way — not just a one-off trick for a single prompt. Why worry? Because it could let the model generate instructions or content that enable harm in biological contexts.
Think of it like finding a master key for a lab door instead of a single loose screw: one key opens many rooms. Detecting and rewarding these discoveries is practical: instead of waiting for someone to misuse the model in the real world, OpenAI invites experts to try to break it under controlled rules. Sound sensible? Better to find the holes in a safe environment than in the wrong hands.
How to apply and who can participate
The application is short: name, affiliation, and experience. OpenAI reviews submissions continuously.
Key requirements: an active ChatGPT account and willingness to sign an NDA to access the private platform.
If you already applied to the GPT-5.5 Bio Bounty, you don’t need to sign up again; you’ll be considered for the program’s continuation.
If you work in security, biosafety, or AI and have experience testing model limits, this is a concrete chance to collaborate with an institution that pays you to find flaws before they become problems. Want an everyday example? Imagine being paid to stress-test a smart home so it won’t accidentally open doors — here you’re doing the same, but for models that could affect biological safety.
And after? Other related programs
OpenAI also reminds you there are other public bounty programs: the Safety Bug Bounty and the Security Bug Bounty, aimed at different types of vulnerabilities. If your interest goes beyond biosafety, it’s worth checking those programs to find the best fit for your skills.
This move from OpenAI shows something simple but important: AI safety is no longer optional or an afterthought. It’s an active practice that involves the community, pays for results, and tries to anticipate risks before they cause harm. If you work at the intersection of AI and biology, you have a window to contribute responsibly and get compensated for it.
