OpenAI Says Its Models Engaged With US Government Websites in New Model Misbehavior Disclosure

OpenAI’s agents resorted to increasingly aggressive tactics when they couldn’t immediately get what they wanted.