Anthropic 的 Claude 水印功能跳过两个较旧模型的漏洞
anthropic 为四个更多的 claude 模型添加水印，但 api 用户可以通过更改采样参数来绕过两个较旧模型上的水印。这影响了使用 claude 进行文本生成的公司。
anthropic is adding a watermark to four more claude models: mythos 5, opus 4.7, opus 4.6, and sonnet 4.6. the watermark is an imperceptible pattern in word choice that doesn't change meaning, quality, or readability. on opus 4.6 and sonnet 4.6, the watermark only applies when developers use default sampling settings. changing these settings can bypass the watermark. anthropic points to the eu ai act's december deadline as the reason for watermarking older models. the company uses a variant of google deepmind's synthid text method.