AI
OpenAI Temporarily Blocks Researchers From Cybersecurity Program Due to Technical Glitch
A technical error caused OpenAI to revoke access to its Trusted Access for Cyber program for select researchers, who now must reapply for verification.
TechCrunch — AI
Multiple cybersecurity researchers reported losing access this week to OpenAI's Trusted Access for Cyber (TAC) program, a specialized initiative that allows vetted security professionals to use the company's most capable AI models with reduced content restrictions. When affected researchers opened ChatGPT's cybersecurity portal, they encountered error messages suggesting their identities could not be verified or that their accounts were ineligible.
OpenAI acknowledged the disruption as a technical mistake affecting a limited cohort of users and asked researchers to reapply and go through the verification process again. The company confirmed the issue in a statement but provided limited detail on the scope or cause of the outage.
About the TAC Program#
The TAC initiative exists to balance two competing security needs. By giving legitimate defenders early access to advanced models with fewer safeguards, researchers can more quickly identify and report vulnerabilities to vendors—theoretically accelerating patch deployment. Simultaneously, restricting access aims to keep malicious actors from weaponizing the same tools to discover exploits.
Similar programs exist elsewhere: Anthropic operates a comparable offering called the Cyber Verification Program (CVP). Joining either requires researchers to submit identity documentation and pass company vetting.
Who Was Affected#
TechCrunch confirmed at least five researchers experienced revoked access. Notably, all of them reported being based outside the United States and Europe, hinting that the technical malfunction may have been geographically specific.
OpenAI recently launched Daybreak Blue, its newest TAC tier for individual researchers, on August 10. This access level provides frontier models including GPT-5.6 Sol and is intended for vulnerability discovery, malware analysis, and incident response. A higher-tier option, Daybreak Red, offers even more permissive tooling for offensive security research.
The Broader Tension#
The incident underscores ongoing friction between AI safety guardrails and legitimate security work. In recent months, both defensive and offensive researchers have publicly criticized the restrictive default settings that Anthropic and OpenAI have imposed, arguing that overly cautious filters obstruct essential defensive and exploratory work.
