OpenAI glitch revokes vetted cyber researchers' access — and its own recovery process won't let some back in

A technical issue in OpenAI's Trusted Access for Cyber (TAC) program wiped the approved status of security researchers who had already passed identity verification, resetting their accounts to "Start verification" as though they had never been cleared. The Daybreak Blue access tier, which gives vetted defenders expanded capabilities in Codex Desktop and the CLI, also vanished from their accounts without warning.

OpenAI told affected users to re-verify, but for some the system now says their accounts are "ineligible" — and support has confirmed it cannot reset the verification state, restore previous approval, or override the new decision. OpenAI declined to answer The Register's questions about how many researchers were affected, what caused the removals, or why previously cleared users are now failing re-verification.