Gpt

OpenAI Bio Bug Bounty
Testing for universal jailbreaks for biorisks
As part of our ongoing efforts to strengthen our safeguards for advanced AI capabilities in biology, we’re evolving our GPT‑5.5 Bio Bug Bounty to become an ongoing private program—the OpenAI Bio Bounty Program. The program will remain focused on universal jailbreaks that can defeat our predefined biosafety challenge against OpenAI’s frontier models, starting with GPT‑5.6 and going forward.
We’re also excited to announce that we’re increasing reward amounts. The reward for a universal jailbreak for the OpenAI Bio Bounty Program has been raised from $25,000 to $50,000 for both GPT‑5.6 and GPT‑5.5. Smaller awards may be granted for partial wins at our discretion.
We will continue to honor the scope for the GPT‑5.5 Bio Bounty Program as originally announced, with testing ending on July 27, 2026. After this date, only GPT‑5.6 will be in scope; we will continue to communicate any future changes in scope to our researchers.
Interested applicants should apply through our rolling application process(opens in a new window). Once selected, accepted applicants will be onboarded to the bio bug bounty platform. All past applicants to the GPT‑5.5 Bio Bounty Program will not need to reapply.
Submit a short application here(opens in a new window) (name, affiliation, experience). Accepted applicants must have existing ChatGPT accounts to apply, and will sign a NDA. Apply now and help us make frontier AI safer.
If you’re interested in supporting OpenAI’s work to deliver safe and secure artificial intelligence beyond the Bio Bounty program, you can learn about our Safety Bug Bounty(opens in a new window) and Security Bug Bounty(opens in a new window) programs.

Facts Only

* The GPT-5.5 Bio Bounty is evolving into the ongoing OpenAI Bio Bounty Program.
* The focus remains on testing universal jailbreaks against predefined biosafety challenges for frontier models, starting with GPT-5.6 and onward.
* The reward for a universal jailbreak has increased from $25,000 to $50,000 for GPT-5.6 and GPT-5.5.
* Testing for the GPT-5.5 Bio Bounty Program will conclude on July 27, 2026.
* After July 27, 2026, only GPT-5.6 will be in scope.
* Applicants must apply through a rolling process and must possess existing ChatGPT accounts to apply.
* Accepted applicants will be onboarded to the bio bug bounty platform and sign an NDA.
* Past GPT-5.5 applicants do not need to reapply.

Executive Summary

OpenAI is evolving the Bio Bug Bounty to become an ongoing private program called the OpenAI Bio Bounty Program, focusing on universal jailbreaks against their frontier models starting with GPT-5.6 and moving forward. The reward for a universal jailbreak has been increased from $25,000 to $50,000 for both GPT-5.6 and GPT-5.5, with potential smaller awards for partial wins. The scope of the GPT-5.5 Bio Bounty Program remains as originally announced, with testing concluding on July 27, 2026. After that date, only GPT-5.6 will remain in scope, with future scope changes to be communicated. Applicants can join via a rolling application process and must have existing ChatGPT accounts, signing an NDA upon acceptance. Additional opportunities exist through OpenAI's Safety Bug Bounty and Security Bug Bounty programs for supporting AI safety efforts.

Full Take

The evolution of the Bio Bounty into a persistent private program signals a shift from episodic testing to continuous adversarial defense against advanced biological AI capabilities, suggesting that vulnerability assessment in frontier models is now viewed as an ongoing security imperative rather than a discrete event. The increased reward structure ties incentivization directly to achieving broader jailbreaks across multiple model versions, potentially streamlining the discovery of fundamental failure modes across the GPT-5 and subsequent generations. The sunsetting of the initial scope for GPT-5.5 in favor of focusing future efforts on GPT-5.6 suggests a deliberate refinement of focus toward the most current and advanced systems while maintaining a long-term commitment to safety research. This creates an architectural tension: balancing the archival testing requirements of past programs with the immediate security concerns of emergent models. The emphasis on applying for existing ChatGPT accounts acts as a gate, potentially prioritizing actors already engaged in the AI ecosystem, which raises questions about accessibility and whether this structure inadvertently reinforces existing power dynamics within the safety research community. What role does temporal demarcation play in the long-term strategy of ensuring that safety testing keeps pace with the rapid evolution of these models?

Sentinel — Human

Confidence

The text exhibits the direct, procedural tone of a formal announcement, suggesting it originated from an internal or official source rather than general journalistic reporting.

Signals Detected

Moderate sentence length variance and direct, transactional tone.

Clear focus on program mechanics and scope changes; lacks overt emotional layering.

Standard corporate announcement structure; lacks verbatim template matching.

Specific dates (July 27, 2026) and concrete monetary increases suggest factual reporting of internal policy changes.

Human Indicators

The structure reads like an official press release detailing program evolution, which aligns with typical corporate communication practices.