OpenAI Blames System Error for Revoking Researchers’ Cyber Access

OpenAI’s Trusted Access for Cybersecurity, or TAC, program is designed to give vetted security researchers access to advanced AI models under fewer safeguards than ordinary users face. The initiative is intended to help defenders identify and repair software vulnerabilities before they can be exploited. Reliable access is critical because abrupt restrictions can interrupt testing and weaken confidence in partnerships between AI developers and the cybersecurity community.

Several researchers participating in TAC said their permissions were suddenly revoked, leaving them unable to use the less restricted, advanced models for security testing. OpenAI later acknowledged that the removals were caused by a system error rather than an intentional change to the program. The company did not disclose how many researchers were affected, the exact date the error began or when full access would be restored.