Event ·Multi-source

安全研究人员用Anthropic的Claude在72小时内入侵OpenAI内部系统

三名安全研究人员利用Anthropic的Claude模型，通过OpenAI社区论坛在72小时内成功入侵其内部系统。Opus 5版本成功绕过了其前代模型无法突破的常见安全措施，表明新一代AI模型能大幅缩短利用安全漏洞所需的时间和专业知识。

Current read

三个安全研究员用Claude在72小时内就攻破了OpenAI的内部系统，这个案例很具体，能让你了解AI模型在安全领域的实际能力。

11 sources78° AI heatLast updated 2026/9/18 17:20:30

Evidence

11 sources

Primary sourceDecoder

2026/9/18 17:20:30View source

Related source 1theverge

2026/9/18 15:30:16View source

Related source 2lmarena.ai

2026/9/16 19:13:56View source

Related source 3techcrunch

2026/9/18 14:00:14View source

Related source 4宝玉

2026/9/18 04:57:51View source

Related source 5IT之家

2026/9/18 05:26:43View source

Related source 6Gary Marcus

2026/9/18 21:58:14View source

Related source 7AI Will

2026/9/20 08:54:06View source

Related source 8Genk

2026/9/20 11:01:00View source

Related source 9Justine Moore

2026/9/16 23:48:07View source

Related source 10arXiv: OpenAI

2026/9/17 02:14:53View 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