Event ·Multi-source

OpenAI将限制Astra模型网络安全功能

OpenAI宣布将推出首个达到'关键'网络安全能力门槛的模型Astra。该模型能无需人类干预，在多个安全系统中识别并开发零日漏洞利用程序。OpenAI将采取分级开放策略，仅向小批测试人员开放高级网络安全任务，并通过Daybreak Blue计划向更广泛用户开放防御性用途。此次发布吸取了Hugging Face事件教训，强化了安全护栏。

Current read

OpenAI的Astra模型能自主发现零日漏洞，但为防滥用将限制其网络安全功能，只开放给防御性用途。

11 sources73° AI heatLast updated 2026/9/1 23:40:04

Evidence

11 sources

Primary sourceIT之家

2026/9/1 23:40:04View source

Related source 1OpenAI Blog

2026/9/1 13:00:00View source

Related source 2techcrunch

2026/9/1 21:06:24View source

Related source 3VTV Công nghệ

2026/9/3 07:43:00View source

Related source 4Decoder

2026/9/3 19:25:40View source

Related source 5Amjad Masad

2026/8/31 00:08:36View source

Related source 6Gary Marcus

2026/8/31 15:27:55View source

Related source 7Paul Graham

2026/8/31 18:28:34View source

Related source 8Anthropic

2026/9/1 00:07:53View source

Related source 9VnExpress Số hóa

2026/9/2 14:58:03View source

Related source 10宝玉

2026/9/2 18:25:23View 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