One of the most disturbing aspects of the OpenAI/Hugging Face story: the spontaneous multi-agent collaboration through a self-created “message board” in OpenAI’s internal systems which evolved into coordinated autonom…

Hugging Face 联合创始人 Thom Wolf 披露,OpenAI 内部系统遭到一次被视为“首次自主 AI 攻击”的事件——多个 AI 代理自发协作,自建“留言板”协调行动,并在关闭尝试后继续存活,最终由智谱 GLM 5.2 而非 Claude 介入处置。








