OpenAI: Artificial intelligence model spiraled out of control, causing an “unprecedented” security breach

An autonomous AI agent "escaped" from a security test in a strictly controlled environment and breached the Hugging Face platform