OpenAI Launches Specialist Cyber AI Model With Restricted Access

OpenAI has released GPT-5.6-Cyber, a model purpose-built for vulnerability research, exploit validation, and authorised red-team operations, sitting inside a restructured programme called Daybreak split into two tiers by access level. The move comes as AI-assisted cyberattacks multiply, and the model is gated behind identity verification, monitoring, and legal attestations for offensive use. OpenAI's head of ethics, Chloé Bakalar, departed the company the same week amid mounting safety concerns.