OpenAI: Access Revoked to Daybreak Blue, Re-Verification Required

A limited number of users no longer have access to Daybreak Blue, the individual level of the Trusted Access for Cyber device, with OpenAI citing a technical incident and requiring re-verification. Researchers reported outages on Wednesday on the Cyber page of ChatGPT. Launched on August 10, Daybreak Blue provides access to cutting-edge generalist models, including GPT-5.6 Sol, while a higher level, Daybreak Red, is dedicated to cybersecurity research. Those affected reside outside the United States and Europe, and the exact scope of the issue remains unknown.

OpenAI launches Daybreak Blue and Daybreak Red for cybersecurity

OpenAI launched Daybreak Blue on August 10. This level, aimed at individual researchers, provides access to cutting-edge generalist models, including GPT-5.6 Sol, with protections tailored for authorized defensive security work. The company presents it as the recommended starting point for most defenders, supporting vulnerability discovery, secure code review, malware analysis, incident response, and patch validation.

At the same time, OpenAI introduced Daybreak Red, a higher level that provides access to models specifically designed for cybersecurity research. According to the company, it allows verified users to conduct authorized vulnerability research, exploit validations, and security testing.

OpenAI requests re-verification after a technical incident

OpenAI attributes these outages to a technical incident. One researcher reported receiving a message indicating the revocation of their access to Daybreak Blue due to a technical issue affecting a limited number of users, with the company clarifying that it was a malfunction on their side and not the desired experience.

On its official channels, OpenAI explained that a restricted set of users no longer has access to Daybreak Blue and will need to re-verify to retain that access. Other exchanges with support also mention a "recent technical issue." In any case, the company is asking affected individuals to reapply and complete the verification process.

Researchers report revocations, uncertain scope

Several security researchers reported that their access was revoked on Wednesday, with messages on the Cyber page of ChatGPT mentioning unverifiable identity or temporary ineligibility. Five researchers say they were affected; all reside outside the United States and Europe. At this stage, the precise cause of the revocations and the total number of affected users remain unknown, and it is only possible that a regional limitation is at play.

OpenAI's Trusted Access for Cyber (TAC) program gives verified researchers access to advanced models with lighter safeguards compared to the general offering. Enrollment requires the submission of identification and verification by OpenAI. TAC pursues a dual objective: to equip trusted defenders to report vulnerabilities and flaws more quickly, and to prevent malicious actors from exploiting these same models. Anthropic offers a comparable program, the Cyber Verification Program. Meanwhile, researchers—both offensive and defensive—have been contesting the safeguards imposed by Anthropic and OpenAI for months, deeming them restrictive for legitimate work.