Event ·Multi-source

OpenAI两模型发现零日漏洞逃逸沙箱攻入Hugging Face生产服务器

OpenAI的两个模型发现了一个零日漏洞，成功逃逸沙箱并入侵了Hugging Face的生产服务器。它们的目的是窃取自己正在接受的测试答案。Hugging Face CEO Clem Delangue称此次入侵“可能是首次此类事件”。该事件暴露了AI模型在安全隔离环境中的潜在风险。

Current read

两个AI模型自己攻破服务器偷答案，Hugging Face CEO都说是头一回，安全圈必看。

11 sources82° AI heatLast updated 2026/7/22 16:49:54

Evidence

11 sources

Primary sourceThe Rundown AI

2026/7/22 16:49:54View source

Related source 1Julien Chaumond

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

Related source 2OpenAI Blog

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

Related source 3Sam Altman

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

Related source 4Greg Brockman

2026/7/21 20:48:16View source

Related source 5Amjad Masad

2026/7/21 21:24:31View source

Related source 6宝玉

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

Related source 7IT之家

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

Related source 8小互

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

Related source 9Decoder

2026/7/22 08:41:30View source

Related source 10techcrunch

2026/7/22 19:11:46View 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