🧬 OpenAI Bio Bug Bounty 2026 — Private rolling program for universal biosafety jailbreak research with rewards up to $50,000.
OpenAI Bio Bug Bounty 2026 Overview
OpenAI has expanded its earlier GPT-5.5 Bio Bug Bounty into an ongoing private initiative called the OpenAI Bio Bounty Program. The program invites selected researchers to test whether a single universal jailbreak can defeat a predefined biosafety challenge across OpenAI frontier models.
The updated program begins with GPT-5.6 while continuing to honor the existing GPT-5.5 testing scope until 27 July 2026. After that date, GPT-5.6 will remain in scope unless OpenAI announces further changes. The maximum reward for a successful universal jailbreak has increased from $25,000 to $50,000, with smaller discretionary awards possible for partial successes.
Quick Overview
| Category | Details |
|---|---|
| Program Name | OpenAI Bio Bounty Program |
| Organizer | OpenAI |
| Program Type | Ongoing private safety research and bug bounty program |
| Main Challenge | Identify a universal jailbreak that defeats a predefined biosafety challenge |
| Models in Scope | GPT-5.6 and GPT-5.5 during the current transition period |
| Maximum Reward | $50,000 for a qualifying universal jailbreak |
| Partial Awards | Smaller awards may be granted at OpenAI’s discretion |
| Application Method | Short rolling application followed by selection and onboarding |
| Required Account | Existing ChatGPT account |
| Confidentiality | Accepted participants must sign an NDA |
About the OpenAI Bio Bounty Program
The Bio Bounty Program is designed to help strengthen safeguards around advanced AI capabilities in biology. Rather than operating as a fully public challenge, it uses a private, application-based model so selected researchers can test high-impact safety scenarios under controlled conditions.
The program focuses specifically on universal jailbreaks. In practical terms, researchers are asked to identify a single prompt or technique that can reliably overcome the program’s predefined biosafety challenge rather than finding a narrow or one-off failure.
Reward Structure
| Outcome | Potential Reward | Notes |
|---|---|---|
| Qualifying universal jailbreak | Up to $50,000 | The increased reward applies to both GPT-5.6 and GPT-5.5 within their applicable scope. |
| Partial success | Discretionary smaller award | OpenAI may grant smaller rewards for partial wins depending on the quality and impact of the finding. |
| Non-qualifying submission | No guaranteed award | Participation or submission alone does not guarantee payment. |
Who Can Apply?
- AI red teamers: Researchers with practical experience testing model safeguards, prompt robustness or adversarial behavior may be suitable applicants.
- Security researchers: Professionals familiar with responsible vulnerability research, controlled testing and documented disclosure may be well aligned.
- Biosafety specialists: Applicants with relevant expertise in biological risk, biosecurity, biotechnology or related safety disciplines may be considered.
- Interdisciplinary researchers: Candidates combining AI safety, security and biological-domain knowledge may be particularly relevant to the challenge.
- Existing ChatGPT users: Accepted applicants must already have a ChatGPT account before participating.
- NDA-ready participants: Selected researchers must be willing to sign a non-disclosure agreement before onboarding.
Application Requirements
- Name: Applicants must provide their full name through the short application form.
- Affiliation: Candidates should mention their university, research organization, company or independent-research status.
- Experience: The application should briefly explain relevant AI red-teaming, security, safety or biosafety experience.
- ChatGPT account: An existing account is required for accepted applicants to participate.
- Confidentiality agreement: Selected participants must complete the required NDA before gaining program access.
How to Participate
- Open the application portal: Use the official OpenAI application page linked below.
- Complete the short form: Provide your name, affiliation and concise summary of relevant experience.
- Submit for review: Applications are reviewed through a rolling process rather than one fixed public deadline.
- Wait for selection: Only accepted applicants will be invited to continue with onboarding.
- Complete onboarding: Selected participants must have a ChatGPT account and sign the required NDA.
- Access the platform: Accepted researchers will be onboarded to the private bio bug bounty platform.
- Conduct authorized testing: Participants should follow the provided scope, platform rules and responsible-disclosure requirements.
- Submit findings: Eligible discoveries should be documented and submitted through the prescribed bounty workflow.
GPT-5.5 and GPT-5.6 Scope Timeline
| Model | Status | Important Date |
|---|---|---|
| GPT-5.5 | Temporarily remains in scope under the previously announced program terms | Testing ends on 27 July 2026 |
| GPT-5.6 | Current frontier model in scope for the ongoing Bio Bounty Program | Continues after the GPT-5.5 transition |
| Future frontier models | May be added or updated through future scope announcements | Researchers will be informed of changes |
Important Participation Conditions
- Private access: Applying does not automatically grant access; participation depends on selection and onboarding.
- Rolling applications: Interested researchers may submit applications through the continuing intake process.
- No reapplication for past applicants: Previous applicants to the GPT-5.5 Bio Bounty Program do not need to apply again.
- Controlled testing: Research must remain within the authorized program scope and platform instructions.
- Confidentiality: Findings, prompts, outputs and program communications may be covered by the participant NDA.
- Reward discretion: OpenAI determines whether a submission qualifies for the full or partial award.
Practical Application Tips
- Explain relevant experience clearly: Focus on AI red teaming, model evaluation, cybersecurity, biosafety or controlled adversarial testing.
- Use concrete examples: Briefly mention past research projects, competitions, publications or responsible-disclosure work where appropriate.
- Keep the application focused: The form is short, so prioritize information directly connected with the program’s safety-testing objectives.
- State your affiliation accurately: Independent researchers should identify themselves honestly rather than inventing an organizational connection.
- Review confidentiality obligations: Be prepared to follow NDA and disclosure rules before participating.
- Do not test outside authorization: Wait for acceptance and formal onboarding before attempting any program-specific research.
Common Mistakes to Avoid
- Treating the application as automatic acceptance: The program is private and requires OpenAI selection.
- Submitting vague experience: Generic statements without relevant red-team, safety, security or biosafety context may weaken an application.
- Ignoring the NDA requirement: Accepted participants must agree to confidentiality terms before accessing the platform.
- Assuming every finding earns $50,000: The maximum reward applies only to qualifying universal jailbreaks.
- Reapplying unnecessarily: Past GPT-5.5 Bio Bounty applicants do not need to submit another application.
- Testing without authorization: Researchers should remain within the official program scope and onboarding process.
Frequently Asked Questions
What is the OpenAI Bio Bounty Program?
It is an ongoing private research program focused on identifying universal jailbreaks that can defeat predefined biosafety challenges in frontier AI models.
What is the maximum reward?
The maximum announced reward is $50,000 for a qualifying universal jailbreak.
Which models are currently in scope?
GPT-5.6 is the ongoing model in scope, while GPT-5.5 remains temporarily included until 27 July 2026.
Are partial rewards available?
Yes. OpenAI may grant smaller awards for partial wins at its discretion.
Is the program public?
No. It is a private program that requires application, selection and onboarding.
Who should apply?
Researchers with experience in AI red teaming, security, AI safety, biosafety or related fields may apply.
Is a ChatGPT account required?
Yes. Accepted applicants must have an existing ChatGPT account.
Is an NDA required?
Yes. Selected participants must sign a non-disclosure agreement before participation.
Do previous GPT-5.5 applicants need to reapply?
No. OpenAI states that past applicants do not need to submit another application.
Is there one fixed application deadline?
The application process is rolling, so applicants may submit through the ongoing intake process.
Final Guidance
The OpenAI Bio Bug Bounty 2026 may suit experienced AI red teamers, cybersecurity researchers and biosafety specialists interested in authorized testing of frontier-model safeguards.
Applicants should submit a concise rolling application, describe their relevant experience accurately and wait for formal acceptance before conducting any program-specific testing.
🧬 OpenAI Bio Bug Bounty 2026 | Rewards Up to $50,000 | Rolling Applications.
📝 Apply Now 🌐 View Official Announcement