Several cybersecurity researchers say OpenAI suddenly revoked their access to the company’s Trusted Access for Cyber program, which provides vetted defenders with frontier AI models stripped of some guardrails so they can discover and report software vulnerabilities.
Multiple researchers posted on OpenAI’s official support forums and on X on Wednesday, reporting that when they tried to access the ChatGPT Cyber page, a message appeared saying their identity could not be verified or that their account was “ineligible at this time.” The affected researchers told TechCrunch they had received no prior warning of the change.
OpenAI Blames Technical Error
OpenAI attributed the disruption to a software bug. In an email shared with TechCrunch, the company told one affected researcher that their access to Daybreak Blue, the highest vetted tier of the TAC program, was revoked “due to a technical issue affecting a limited number of users.”
This was an issue on our end, and not the user experience we want to deliver.
In a follow-up post on its developer accounts channel, OpenAI confirmed that a “limited set of users’ access to Daybreak Blue is no longer active” and asked them to re-verify their identity. All five researchers TechCrunch spoke to said they live outside the United States and Europe, raising questions about whether the revocations were region-linked.
What Daybreak Blue and Red Actually Do
The Trusted Access for Cyber program, launched in February 2026 as a successor to OpenAI’s earlier cybersecurity grant program, has two tiers. Daybreak Blue, which uses GPT-5.6 Sol, is the recommended tier for most defenders and supports vulnerability discovery, secure code review, malware analysis, and incident response. Daybreak Red, using the specialized GPT-5.6 Cyber model, is reserved for authorized penetration testing and exploit validation.
Anthropic runs a parallel program called the Cyber Verification Program. Both programs aim to give security professionals access to AI capabilities that might otherwise be blocked by safety filters, while keeping the same models locked down for the general public.
A Growing Friction Between AI Safety and Cybersecurity
The revocations add to a simmering tension between AI companies and the security community. In recent months, both defensive and offensive researchers have complained that guardrails imposed by Anthropic and OpenAI prevent legitimate security work, including vulnerability research and red-team exercises.
The issue is particularly acute as AI models become more capable of finding software flaws. Anthropic warned earlier this year that its Mythos model could identify vulnerabilities in ways potentially weaponizable by malicious actors, prompting White House scrutiny and temporary access restrictions. Security researchers argue that restricting access for defenders while offensive capabilities continue to improve leaves organizations more exposed.
OpenAI has not said how many users were affected or whether the re-verification process will fully restore access to Daybreak Blue and Red tiers. The company did not respond to questions about whether the revocations were tied to geographic restrictions.
Sources: TechCrunch; OpenAI developer forums; OpenAI Trusted Access for Cyber documentation
discussion