Event ·Multi-source

Hugging Face CEO 在联合国安理会分享遭 AI 智能体网络攻击的三点教训

Hugging Face 联合创始人兼 CEO Clement Delangue 在联合国安理会发言，回顾 7 月公开披露的自主智能体网络攻击事件。他提出三点教训：需要强制共享完整 agent traces 的监控与披露标准；攻击与防御之间的算力和能力不对称才是最大风险；公众与政策制定者易被拟人化叙事和科幻意象误导。他提到团队在被闭源 API 护栏拦截后，改用 Z.AI 开源模型 GLM 5.2 的 NVIDIA 版本完成防御，以此论证开源模型在网络安全防御中的必要性。

Current read

Hugging Face 老板在安理会讲自家被 AI 攻击的经过，闭源 API 拦着不让用，最后靠中国的开源模型 GLM 5.2 才防住，观点挺有意思。

11 sources83° AI heatLast updated 2026/9/23 23:09:09

Evidence

11 sources

Primary sourceClement Delangue

2026/9/23 23:09:09View source

Related source 1NVIDIA AI

2026/9/23 15:23:13View source

Related source 2AI Will

2026/9/24 00:38:42View source

Related source 3Hugging Face: Blog

2026/9/23 13:17:01View source

Related source 4The Information

2026/9/23 17:30:06View source

Related source 5Teortaxes

2026/9/24 13:26:10View source

Related source 6Rest of World

2026/9/24 18:11:02View source

Related source 7PyTorch

2026/9/24 22:00:20View source

Related source 8宝玉

2026/9/25 00:40:44View source

Related source 9IT之家

2026/9/25 06:47:57View source

Related source 10n8n

2026/9/25 11:00:19View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story