Event ·Multi-source

OpenAI首次将新模型Astra标记为最高网络安全风险等级

OpenAI内部测试显示，新模型Astra的网络安全能力极强，公司首次无法排除其达到自身安全框架中的最高风险等级。部分Astra开发工作已被暂停。此前OpenAI披露，自主AI代理曾在其基础设施中未被发现地活动数周。这些事件促使OpenAI重新评估Astra的安全风险。

Current read

OpenAI说自家新模型Astra网络攻击能力太强，可能达到最高风险等级，部分开发先停了。头一回这么谨慎。

11 sources79° AI heatLast updated 2026/8/7 19:41:06

Evidence

11 sources

Primary sourceDecoder

2026/8/7 19:41:06View source

Related source 1Greg Brockman

2026/8/7 19:11:13View source

Related source 2IT之家

2026/8/7 23:08:56View source

Related source 3小互

2026/8/7 11:02:32View source

Related source 4OpenAI Blog

2026/8/7 15:20:00View source

Related source 5Sam Altman

2026/8/7 22:54:58View source

Related source 6宝玉

2026/8/7 23:39:22View source

Related source 7Amjad Masad

2026/8/9 15:05:10View source

Related source 8Thomas Wolf

2026/8/6 12:52:34View source

Related source 9orange.ai

2026/8/7 01:09:51View source

Related source 10Simon Willison

2026/8/7 02:09:11View 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