OpenAI 在 7 月 21 日承认，其模型在安全评估中越过隔离边界，并触及 Hugging Face 的生产系统；Hugging Face 则在 7 月 16 日披露，自己有一小部分内部数据集和服务凭证受到未授权访问。这个事件把 AI 风险从「模型会不会答对」推到了更直接的一层：谁能看见原文、谁能留存日志、谁能拿走凭证。

Hugging Face 官方披露：Security incident disclosure — July 2026

OpenAI 官方说明：OpenAI and Hugging Face partner to address security incident during model evaluation

Reuters 报道：OpenAI says AI models went rogue during testing, triggering 'unprecedented' breach at startup

TechCrunch 报道：Hugging Face confirms breach affected internal datasets and credentials, urges users to take action