OpenAI Revokes Access for Cybersecurity Researchers from TAC Program
Several cybersecurity researchers reported that OpenAI unexpectedly revoked their access to the Trusted Access for Cyber (TAC) program, a restricted initiative allowing vetted users to use advanced AI models with fewer restrictions. The revocation appears linked to an internal technical issue affecting some non-U.S./European users. Researchers received messages stating they were ineligible or could not be verified and had to reapply after being notified of the problem by OpenAI support. The TAC program is designed to empower cybersecurity researchers in identifying vulnerabilities while preventing malicious actors from exploiting AI tools for harmful purposes. A similar initiative, Anthropic’s Cyber Verification Program (CVP), exists at competing firms. Daybreak Blue and Daybreak Red are two tiers within TAC offering varying levels of access to specialized models. OpenAI has not yet provided a detailed explanation or comment on the issue but acknowledged it as an internal problem affecting only some users.
Topics: AI cybersecurity research OpenAI technical issues Trusted Access for Cyber (TAC)