OpenAI cyber-research access revoked for some vetted users due to error

Multiple cybersecurity researchers reported on OpenAI's official support forums and on X this past Wednesday that their access to the Trusted Access for Cyber (TAC) program had been suddenly revoked. When they tried to open ChatGPT's Cyber page, they were met with messages saying their identity could not be verified or that their account was currently ineligible.

TAC allows vetted security researchers to use OpenAI's most capable AI models with fewer safety restrictions than standard accounts, in order to help them find and report vulnerabilities more effectively. Anthropic runs a comparable initiative called the Cyber Verification Program. Both programs aim to help trusted defenders get flaws patched faster while keeping the same powerful models out of the hands of malicious actors. Access to TAC requires researchers to submit identification and pass OpenAI's vetting process.

It remains unclear exactly why the access was revoked or how many users were affected. TechCrunch spoke with five researchers who experienced the problem. One received an email stating that access to Daybreak Blue, TAC's newest vetted tier, had been revoked because of a technical issue affecting a limited number of users. The message acknowledged that this was not the intended user experience and asked the researcher to reapply and complete verification again. Another researcher was told the same by OpenAI support after raising the issue on the company's forum.

All five researchers TechCrunch interviewed said they are based outside the United States and Europe, suggesting the revocations may be concentrated in certain regions. OpenAI, responding via a public post on X, said a limited set of users' access to Daybreak Blue was no longer active and that they would need to re-verify to regain it.

Daybreak Blue, launched on August 10, is aimed at individual researchers and provides access to frontier general-purpose models, including GPT-5.6 Sol, with safeguards designed for authorized defensive security work such as vulnerability discovery, code review, malware analysis, incident response and patch validation. Alongside it, OpenAI introduced a higher tier called Daybreak Red, built specifically for authorized vulnerability research, exploit validation and security testing. In recent months, both defensive and offensive security researchers have criticized the guardrails imposed by OpenAI and Anthropic, saying they hinder legitimate research work.