通过请求 Base64 响应,轻松绕过文本水印。

2 分•作者: hanneshdc•大约 1 个月前•原帖
Anthropic计划推出基于SynthID的文本水印技术,该技术以确定性方式调节令牌选择机制背后的伪随机数生成器(PNRG),使得后续的工具能够估计文本是否由人工智能撰写。 要绕过这一措施,目前的形式非常简单: 1. 请求以base64编码的响应/论文/电子邮件。 2. 使用 https://www.base64decode.org/ 解码。 虽然编码后的文本会遵循水印分布,但解码后的文本则不会。
查看原文
Anthropic plans to release text watermarking based on SynthID, which modulates the PNRG behind the token choosing mechanism deterministically in a way that lets another tool later estimate whether text was written by AI.<p>This is trivial to bypass that, it&#x27;s current form: 1. Ask for the response&#x2F;essay&#x2F;email encoded in base64 2. Decode it using https:&#x2F;&#x2F;www.base64decode.org&#x2F;<p>While the encoded text will follow the watermarked distribution, the decoded text will not.