Event ·Multi-source

安全研究人员用 Claude Opus 5 攻破 OpenAI 员工账户

3名安全研究人员利用Claude Opus 5的图片上传漏洞，成功攻破OpenAI员工账户。他们随后让该账户中的Codex在OpenAI内部代码库提交了一个PR。整个攻击的Token成本不到3000美元。Opus 4.8难以完成此利用，但Opus 5发布后仅几小时就成功突破。

Current read

朋友发现用Claude Opus 5能攻破OpenAI员工账户，成本还不到3000美元，这挺有意思的。

11 sources78° AI heatLast updated 2026/9/20 08:54:26

Evidence

11 sources

Primary sourceAI Will

2026/9/20 08:54:26View source

Related source 1Blognone

2026/9/22 01:30:57View source

Related source 2techcrunch

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

Related source 3theverge

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

Related source 4Decoder

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

Related source 5Genk

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

Related source 6The Rundown AI

2026/9/18 10:30:19View source

Related source 7Gary Marcus

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

Related source 8掘金本周最热

2026/9/21 09:24:25View source

Related source 9Simon Willison’s Weblog

2026/9/21 23:09:20View source

Related source 10shao__meng

2026/9/21 23:44:32View 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