Beat GPT-5.5's Bio Safeguards and Win $25,000!

Hey everyone, it's Shii-chan! Today's news is all about poking holes in AI safety on purpose, which is oddly exciting.

OpenAI News openai.com

What was announced?

OpenAI just kicked off a Bio Bug Bounty for GPT-5.5, announced on OpenAI's News. They're inviting researchers with experience in AI red teaming, security, or biosecurity to try to break the model's biology-related safeguards on purpose.

It's a serious call for people who can hunt for real weaknesses.

Why it matters

Powerful AI models need strong guardrails around sensitive biology topics, but the only way to know if those guardrails really hold is to have outside experts attack them hard.

By inviting trusted red-teamers to look for weaknesses, OpenAI gets a much more honest picture of how safe GPT-5.5 actually is.

What changes

The challenge is easy to state but hard to pull off. Your goal is to find one universal jailbreak prompt that answers all five bio safety questions from a clean chat, without triggering moderation. The model in scope is GPT-5.5 in Codex Desktop only.

The first person to land a true universal jailbreak that clears all five questions gets $25,000. Smaller awards may also be granted for partial wins, at OpenAI's discretion.

Dive Deep

Here's the schedule. Applications open on April 23, 2026 with rolling acceptances and close on June 22, 2026. Testing begins on April 28, 2026 and ends on July 27, 2026.

Access is by application and invite. OpenAI will send invitations to a vetted list of trusted bio red-teamers while also reviewing new applications. Once selected, you get onboarded to the bio bug bounty platform.

To apply, submit a short application with your name, affiliation, and experience through the application form by June 22, 2026. You'll need an existing ChatGPT account, and you'll sign an NDA. All prompts, completions, findings, and communications are covered by that NDA, so nothing gets shared publicly.

Wrap-up

• OpenAI announced a Bio Bug Bounty for GPT-5.5, scoped to GPT-5.5 in Codex Desktop
• The goal is one universal jailbreak that clears a five-question bio safety challenge from a clean chat, without triggering moderation
• $25,000 for the first true universal jailbreak, plus discretionary smaller awards for partial wins
• Applications run April 23 to June 22, 2026; testing runs April 28 to July 27, 2026, invite- and selection-based, with an NDA required

This one is for you if you have chops in AI red teaming or biosecurity, and for anyone curious about how AI safety actually gets stress-tested!