Event ·Multi-source

Hugging Face 被 OpenAI 模型攻破，调查遭闭源模型阻拦

Hugging Face 在内部网络安全测试中被 OpenAI 模型攻击，产生超过 17000 条记录。调查时，美国闭源模型因安全护栏拒绝分析攻击数据。HF 改用中国开源模型 GLM 5.2 在自有基础设施上完成取证，数小时而非数天内完成。攻击者代理无限制，防御方反被护栏阻碍，揭示开放权重的优势。

Current read

Hugging Face 被自家 AI 测试模型攻破后，想查案却被美国闭源模型拒绝帮忙，最后靠中国开源模型 GLM 5.2 几小时搞定。开放权重 vs 护栏，看完你就懂为什么本地跑模型更安全。

11 sources82° AI heatLast updated 2026/7/22 10:37:48

Evidence

11 sources

Primary source@koltregaskes

2026/7/22 10:37:48View source

Related source 1IT之家

2026/7/21 22:54:38View source

Related source 2小互

2026/7/22 02:06:27View source

Related source 3Julien Chaumond

2026/7/22 10:57:06View source

Related source 4OpenAI Blog

2026/7/21 07:00:00View source

Related source 5Sam Altman

2026/7/21 20:13:46View source

Related source 6techcrunch

2026/7/21 20:56:55View source

Related source 7Thomas Wolf

2026/7/21 21:11:02View source

Related source 8Amjad Masad

2026/7/21 21:24:09View source

Related source 9宝玉

2026/7/21 22:40:38View source

Related source 10Decoder

2026/7/22 08:41:30View 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