OpenAI releases new model that it says triggered internal security measures

The AI giant says its model better follows user intent and crosses a frontier for cybersecurity capabilities and could be artificial general intelligence.