OpenAI is also expanding in its efforts to enhance AI safety by offering security researchers a chance to earn up to $50,000 for identifying critical vulnerabilities in its new AI model GPT-5.6.
The company officially expanded its Bio Bug Bounty Program to include GPT-5.6 as well, raising the maximum amount of payments to $50,000 from the $25,000 that was given during the GPT-5.5 testing program. The initiative is to identify the weaknesses of the model's biology-based safety protections in advance of them being exploited by malicious actors.
As AI systems develop so fast, there are concerns in some domains such as biology, chemistry, and cybersecurity that they might be weaponized. OpenAI says that early detection of vulnerabilities is essential in order to make AI models and their applications safe.
Focus on "Universal Jailbreaks"
Unlike the usual prompt engineering techniques that may temporarily bypass an AI model's restrictions, OpenAI is seeking "universal jailbreaks".
Those are methods that can consistently circumvent GPT-5.6's built-in biosafety safeguards across multiple conversations and scenarios.
Exploring these persistent vulnerabilities will allow engineers to improve future safety systems and avoid the potential for misuse.
Researchers who uncover less severe problems might also receive partial rewards, with compensation based on the quality, originality and impact of their findings.
Why OpenAI increased the Reward
OpenAI says GPT-5.6 is its newest generation of AI technology, making comprehensive safety testing more important than ever.
To get broader participation from global cybersecurity and AI research communities, it has doubled the maximum payout to $50,000.
The GPT-5.6 Bio Bug Bounty Programme will replace the GPT-5.5 version after July 27, 2026, when testing of the older model is complete.
Researchers already in the GPT-5.5 program will not need to submit a new application, so they can easily transition into the updated initiative.
Who can apply?
OpenAI has also opened applications to researchers interested in evaluating GPT-5.6's safety systems.
The information of the company’s Bio Bug Bounty page should be submitted to applicants through a Bio Bug Bounty page which must be submitted along with the name, organization and/or institution or institutional affiliation, and research experience, and the specific research experience of the applicant.
Participants are also required to:
Have an active ChatGPT account. Sign a NDA. Before accessing the specialized testing environment, receive approval.
Once accepted, researchers will have access to GPT-5.6 after the GPT-5.5 testing phase officially ends later this month.
It is important to have Bug Bounty Programs.
Bug bounty programs are a cornerstone of cybersecurity and organizations’ ability to identify software vulnerabilities before cybercriminals can exploit them.
OpenAI is doing the same with artificial intelligence by inviting independent experts to rigorously test its latest model under controlled conditions.
That approach is indicative of increasing global interest in AI safety. Governments and regulators are increasingly demanding stronger safeguards as AI systems are increasingly capable of handling more complex science, technical, and medical information.
The biology, cybersecurity and chemical research fields have been particularly important given the potential risks of AI-generated knowledge being compromised if protections are not put in place.
By encouraging independent researchers to stress-test GPT-5.6 before being released, OpenAI is hoping to identify weaknesses early and improve its biosafety protections to improve the overall reliability of future AI systems.
The expanded bug bounty program is emblematic of a larger industry trend: as artificial intelligence gets better, safety and responsible technology development is more and more important as it is the development of AI.
Comments
Please to leave a comment on this article.