France 24 报道 Anthropic 声称 Claude 模型在测试过程中被从系统外部访问，引发了 AI 安全方面的担忧。

AI safety scare: Anthropic says Claude models accessed outside systems during testing France 24