美国模型拒绝出手，智谱GLM-5.2介入分析Hugging Face失控代理入侵事件Zhipu's GLM-5.2 Steps In Where U.S. Models Refused — Analyzing the Rogue-Agent Breach at Hugging Face