OpenAI Revokes Some GPT-5.6 Cyber Access by Error, Forcing Re-Verification
Updated
Updated · TechCrunch · Aug 19
OpenAI Revokes Some GPT-5.6 Cyber Access by Error, Forcing Re-Verification
3 articles · Updated · TechCrunch · Aug 19
Summary
OpenAI told a limited set of researchers their Daybreak Blue access had been revoked because of a technical issue, requiring them to reapply and complete verification again.
Five researchers told TechCrunch they lost entry to the Trusted Access for Cyber program, with ChatGPT showing identity-verification or eligibility errors when they opened the Cyber page.
All five affected researchers were outside the U.S. and Europe, suggesting the mistaken revocations may be concentrated in certain regions, though OpenAI did not say how many users were hit.
Daybreak Blue, launched on August 10, gives vetted defenders access to frontier models including GPT-5.6 Sol with lighter cybersecurity guardrails for vulnerability discovery, malware analysis and incident response.
The disruption lands amid broader complaints from security researchers that OpenAI and Anthropic's guardrails can block legitimate defensive work even as both companies try to keep powerful cyber models from criminals.