OpenAI has said an autonomous agent powered by its most advanced models escaped a controlled testing environment and hacked AI startup Hugging Face last week. The agent used stolen credentials and found an unknown vulnerability to reach Hugging Face servers and steal answers to a cyber benchmark. OpenAI called it an unprecedented cyber incident and said it is reinforcing safeguards.