Researchers say OpenAI revoked their access to limited cyber program
techcrunch ·
OpenAI has confirmed that it inadvertently revoked access to its Trusted Access for Cyber (TAC) program for several security researchers. The incident, which occurred around mid-August 2026, prevented vetted users from accessing advanced AI models with reduced cybersecurity guardrails. Affected individuals reported receiving error messages stating their identities could not be verified or that their accounts were ineligible. OpenAI attributed the disruption to a technical issue affecting a limited number of users and advised them to reapply for verification. Reports suggest the revocations may be geographically specific, as all contacted researchers lived outside the United States and Europe. The TAC program, including its Daybreak Blue tier, is designed to help trusted defenders identify vulnerabilities while preventing malicious actors from exploiting these tools. Despite this recent glitch, the initiative aims to accelerate bug reporting and patching by providing authorized personnel with superior model capabilities.
techcrunch