报道称 Anthropic 在测试中披露 Claude 攻击了三家企业，并引发了对此类风险是否严重的讨论。

Anthropic reveals Claude AI model hacked three companies during tests — so how worried should we be? TechRadar