Anthropic 为 Claude 输出添加水印，批评者质疑文本质量代价
Anthropic's text watermarking for Claude is supposed to make AI-generated content detectable. But critics doubt that word choice stays unaffected, and lawyers are facing new transparency headaches.
回到原文
这里保留短导读，不转载来源网站的完整正文。具体表述和后续更新请以原文为准。
打开 the-decoder.com 原文