Anthropic is offering a clearer look at how it will watermark text produced by its Claude chatbot, responding to a wave of questions from users.
Anthropic published a blog post on Friday that seeks to answer basic questions about its plan to watermark text generated by the Claude chatbot. The post explains how the watermarking process works, whether it can be hidden through editing, and what impact it will have on code produced by the model.
This clarification comes after Anthropic announced earlier in the week that Claude would adopt watermarking in order to comply with the EU AI Act’s Transparency Code, which requires AI companies to use systems that make it possible to identify AI-generated content.
Since the initial announcement, Claude users have been debating the change. The new blog post appears intended to address those discussions directly, providing more concrete details about the technical implementation and its implications for developers who use Claude for coding tasks.
With the blog post, Anthropic is aiming to give users and developers a better understanding of the watermarking system’s practical impact, while affirming its commitment to the transparency requirements of the EU AI Act.
