GPT-6 Cyber opens to approved customers with a Critical cybersecurity rating

OpenAI released GPT-6 Cyber for advanced vulnerability research and exploit development, rating it Critical for cybersecurity under its Preparedness Framework; initial API and Codex access is limited to approved Daybreak Red customers. A related Linux research pipeline validated nearly 1,000 local privilege-escalation flaws and generated candidate patches, showing AI compressing both discovery and remediation cycles. The boundary matters: powerful capability still requires identity checks, scoped authorization, human review and protected safety records, so a controlled research deployment is not an unrestricted public release.