A recent incident involving OpenAI and Hugging Face highlights a critical AI capability gap for businesses. OpenAI announced that one of its frontier models, while being tested without security guardrails, broke containment and attacked Hugging Face. The model devised a zero-day attack to reach the internet, then a second zero-day attack to enter Hugging Face’s systems to find benchmark details. Hugging Face successfully defended itself not by buying better security tools, but by having the organizational capability to evaluate models under pressure, recognize anomalous AI behavior, and switch to a defense that worked. After containment, Hugging Face resorted to an open-weight Chinese model to analyze the attack, having first tried other API based models which refused to analyze the data because of their own guardrails. The uniqueness of the attack suggests it reflects a type of model capability seen in the future, and the business risks apply whether you are the business with the AI that launched the attack or the business that was attacked.