OpenAI Blames Technical Error for Revoking Cyber Researchers' Access to TAC Program
On August 19, 2026, researchers in cybersecurity revealed that OpenAI unexpectedly terminated their access to the Trusted Access for Cyber (TAC) program, which permits approved users to utilize advanced AI models with fewer safety limitations. They received error notifications on the Cyber section of ChatGPT. OpenAI acknowledged that the access revocations were unintentional and resulted from a technical error. TAC is designed to assist trusted defenders in detecting vulnerabilities before they can be exploited. Five researchers affected by this issue, all located outside the U.S. and Europe, reported that their access to Daybreak Blue, which was launched on August 10, was also revoked. OpenAI indicated that re-verification is necessary. Researchers have criticized the limitations set by OpenAI and Anthropic, arguing that these restrictions obstruct valid research.
Key facts
- OpenAI revoked access to the Trusted Access for Cyber (TAC) program for several cybersecurity researchers on or around August 19, 2026.
- OpenAI confirmed the incident was due to an error on its end.
- Affected researchers received messages on ChatGPT's Cyber page saying their identity could not be verified or that their account was ineligible.
- TAC provides vetted researchers access to advanced AI models with fewer cybersecurity guardrails.
- All five researchers interviewed by TechCrunch lived outside the United States and Europe.
- OpenAI asked affected researchers to reapply and complete verification to regain access.
- Daybreak Blue, launched on August 10, grants access to GPT-5.6 Sol and is recommended for defensive security work.
- Anthropic runs a similar program called the Cyber Verification Program (CVP).
Entities
Institutions
- OpenAI
- Anthropic
- TechCrunch
Locations
- United States
- Europe