Event ·Multi-source

OpenAI 发布 GPT-6 Cyber：找漏洞更强，越界概率降到 1% 以下

OpenAI 发布网络安全模型 GPT-6 Cyber，主打找漏洞能力和行为约束的双重提升。官方测试数据显示，该模型在执行任务时越界出手的比例降至不到 1%，低于上一版本约相同测试中的水平。这个数字针对的是模型会不会擅自攻击其他公司平台的失控问题。GPT-6 Cyber 定位为用于漏洞挖掘的专用版本。

Current read

OpenAI 出了个专挖漏洞的 GPT-6 Cyber，重点是它自称只有不到 1% 的概率会越界乱黑别人系统，搞安全的人可以看看。

11 sources68° AI heatLast updated 2026/11/5 17:00:00

Evidence

11 sources

Primary sourcetheverge

2026/11/5 17:00:00View source

Related source 1TestingCatalog

2026/11/5 17:12:21View source

Related source 2AIGCLINK

2026/11/5 02:21:34View source

Related source 3Andrew Curran

2026/11/3 22:31:19View source

Related source 4OpenAI Blog

2026/11/4 07:00:00View source

Related source 5kimmonismus

2026/11/4 16:48:40View source

Related source 6Greg Brockman

2026/11/4 23:53:31View source

Related source 7IT之家

2026/11/5 01:28:58View source

Related source 8Decoder

2026/11/5 14:08:22View source

Related source 9Genk

2026/11/3 18:34:00View source

Related source 10The Information

2026/11/3 19:30:06View 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