Hugging Face Used Chinese AI Model to Analyze Cyberattack After U.S. Model Blocked It

Hugging Face said it used the Chinese open-source model Kimi K3, via Z.ai’s GLM 5.2 running on its own infrastructure, to analyze a cyberattack after safety guardrails on a U.S. frontier AI model prevented it from examining malicious payloads and logs. The company said the incident involved an autonomous AI-driven intrusion and that its security team had to review more than 17,000 logged events to understand what happened. The episode has reignited debate over whether strict AI safeguards can hinder legitimate incident response at the exact moment defenders need tools to inspect attacker behavior.