Several security researchers have reported that their access to OpenAI’s Trusted Access for Cyber (TAC) program has been revoked, leaving them locked out of advanced AI tools used in cybersecurity research.

The issue arose due to a technical error, according to OpenAI. Affected researchers now need to reapply and complete the verification process. The TAC program offers vetted researchers access to more advanced models than those available to regular users, with fewer restrictions.

Five researchers contacted by TechCrunch highlighted this problem, suggesting that it may be limited to certain regions outside of the U.S. and Europe. OpenAI has confirmed that a ‘limited set of users’ are affected, but did not specify how many.

The revocation of access highlights ongoing debates about the balance between security restrictions and the need for researchers to identify vulnerabilities. Both OpenAI and Anthropic offer similar programs aimed at fostering responsible cybersecurity practices while preventing malicious actors from gaining unauthorized access.