The Frontier
OpenAI ships GPT-5.6-Cyber to vetted defenders under Daybreak
Three days after OpenAI said it could not rule out "Critical" cyber capabilities in its next model, it is doing the opposite of slamming the door: shipping a more permissive cyber model to a vetted circle of defenders — and framing it as the only way to close a