OpenAI releases new model that it says triggered internal security measures

The AI giant says its new model better follows user intent and crosses a critical frontier for cybersecurity capabilities.