OpenAI GPT-5.5 Bio Bug Bounty Program Launched

Key Takeaways

• OpenAI has initiated a specialized Bio Bug Bounty program for its GPT-5.5 AI model.
• The program’s core objective is to identify a “universal jailbreak” prompt capable of bypassing biosecurity safeguards within GPT-5.5.
• Only vetted researchers will participate, aiming to prevent potential misuse of advanced AI in biological contexts.
• The initiative highlights a growing intersection between AI security, red teaming, and biosecurity.

OpenAI has rolled out a new Bio Bug Bounty program, specifically targeting its advanced GPT-5.5 artificial intelligence system. This strategic move aims to fortify the AI’s safety protocols and mitigate potential misuse, particularly within the sensitive domain of biology.

The program challenges qualified researchers to determine if GPT-5.5 can be universally “jailbroken” – meaning a single, crafted prompt could bypass its biosecurity protections across multiple scenarios.

At its heart, the initiative focuses on a singular, critical challenge: participants must devise one “universal jailbreak” prompt. This prompt, when entered into a clean chat session, must compel GPT-5.5 to answer all five questions in OpenAI’s biosecurity challenge without triggering any moderation mechanisms.

Strengthening Safeguards for Advanced AI

Essentially, researchers are tasked with discovering whether a meticulously engineered prompt can consistently override the model’s established biological safety guardrails. OpenAI has specified that the testing will be conducted on GPT-5.5 exclusively when running in Codex Desktop.

OpenAI is offering a substantial reward to the first participant who successfully uncovers a genuine universal jailbreak that addresses all five challenge questions. Depending on the outcomes, smaller rewards may also be distributed for partial successes. Applications for this program commenced on April 23, 2026, and will conclude on June 22, 2026.

The testing phase is scheduled to begin on April 28, 2026, and will run until July 27, 2026. Participation is not open to the general public. Instead, OpenAI will extend invitations to a pre-vetted group of trusted bio red-teamers. The company will also review applications from new researchers possessing relevant expertise in AI red teaming, security, or biosecurity.

Prospective participants must submit a brief application form detailing their name, affiliation, and relevant experience. All accepted participants and collaborators are required to possess existing ChatGPT accounts and sign a non-disclosure agreement (NDA). OpenAI has affirmed that all prompts, model outputs, findings, and associated communications will remain strictly confidential under the NDA.

From a broader cybersecurity perspective, this program signifies a growing trend in adversarial testing applied to cutting-edge AI systems. Bug bounty programs have long been instrumental in identifying vulnerabilities across software, cloud platforms, and enterprise solutions. OpenAI is now adapting this proven model to AI safety, inviting experts to actively probe its defenses and pinpoint prompt-based weaknesses before malicious actors can exploit them.

The specific emphasis on biology is particularly crucial, given the potential for powerful AI models to be misused for harmful scientific endeavors if their safeguards prove inadequate. By rigorously testing GPT-5.5 against universal jailbreaks, OpenAI appears to be assessing the resilience of its protective measures under realistic attack conditions.

OpenAI also noted that researchers interested in broader security engagements could explore its existing Safety Bug Bounty and Security Bug Bounty programs. The new GPT-5.5 Bio Bug Bounty program represents an additional layer in this ongoing effort, underscoring the increasing convergence of AI security with biosecurity, red teaming methodologies, and advanced prompt-injection research.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.