Event ·Multi-source

Anthropic 发现 AI 代理也讨厌验证码

Anthropic 的研究显示，其开发的 AI 代理在尝试模拟人类行为时，会自动避开常见的验证码（CAPTCHA）机制。该研究通过模拟 1000 个 AI 代理的上网行为，发现它们会优先选择那些不包含验证码的网站。这一发现揭示了 AI 代理在自动化任务中面临的普遍障碍。

Current read

Anthropic 的研究挺有意思，他们测试了 AI 代理怎么上网，发现它们也讨厌验证码，和人类一样。

11 sources58° AI heatLast updated 2026/9/10 17:54:44

Evidence

11 sources

Primary sourcetechcrunch

2026/9/10 17:54:44View source

Related source 1arXiv: OpenAI

2026/9/9 00:16:53View source

Related source 2IT之家

2026/9/9 11:00:21View source

Related source 3小互

2026/9/9 12:33:38View source

Related source 4The Rundown AI

2026/9/9 17:54:53View source

Related source 5Paul Graham

2026/9/10 08:46:27View source

Related source 6Decoder

2026/9/10 10:31:20View source

Related source 7Thomas Wolf

2026/9/10 10:38:45View source

Related source 8andrew chen

2026/9/10 18:06:16View source

Related source 9Techsauce

2026/9/11 10:36:03View source

Related source 10Simon Willison

2026/9/11 17:49:20View 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