When AI Agents Escaped the Sandbox: What Really Happened in the OpenAI–Hugging Face Incident | AI BriefWire