OpenAI revokes access to cybersecurity research program

The TAC program serves as a sandbox for vetted experts, providing access to advanced AI models with reduced guardrails to assist in malware analysis and incident response. Several affected researchers reported receiving communications from OpenAI attributing the lockout to a technical glitch involving 'Daybreak Blue,' a tier launched on August 10 that grants access to frontier models like GPT-5.6 Sol.

While OpenAI officially characterized the disruption as an internal error affecting a limited group, the impact appears geographically concentrated. Multiple researchers noted they reside outside the U.S. and Europe, fueling speculation about regional policy shifts. The company has instructed those locked out to reapply and complete the identity verification process again to restore their research capabilities. This friction highlights a broader tension between AI developers and the security community, as experts increasingly argue that existing protective guardrails frequently obstruct legitimate vulnerability research and defensive work.