Security researchers approved for OpenAI’s restricted cyber program found themselves locked out this week. The lab says the removals were a technical mistake, not a policy change.
On August 19, complaints surfaced on OpenAI’s support forums and on X. ChatGPT’s Cyber page had begun telling people their identity could not be verified, or that their account was ineligible at the time.
The Trusted Access for Cyber program sits at the center of the mix-up. It hands vetted defenders access to advanced models with fewer guardrails, an arrangement Anthropic copies with its Cyber Verification Program.
TechCrunch spoke to five researchers who lost access. One said an email blamed the change on a technical issue affecting a limited number of users, and stressed the problem was on OpenAI’s end. The company has not said how many accounts were affected or when access will be restored.
