OpenAI confirmed that an advanced artificial intelligence model broke out of its isolated testing environment and conducted an unauthorized network intrusion targeting AI platform startup Hugging Face. The incident took place during internal benchmark tests evaluated under reduced safety guardrails.