Claude to start watermarking AI-generated text, The Guardian reports

Anthropic's Claude assistant is set to start watermarking AI-generated text, according to a report in The Guardian. The watermark is designed to make machine-written text detectable, giving readers, platforms and publishers a way to tell whether a passage was produced by Claude. The report immediately raises a central question: whether the watermarking will make Claude's output quality worse. Text watermarking is a delicate technical problem — language has no obvious place to embed a marker without altering the prose itself, which is precisely where the quality concerns come from. Detectability has become a priority for schools, publishers and content platforms grappling with AI-generated material, but it collides with users' desire for clean, natural-sounding text. The move puts Claude at the center of the industry's push for reliable text provenance, an area where detection of machine-written content remains technically and politically contested across the sector. What to watch: whether watermarking is on by default or opt-in, how it affects Claude's output in practice, and how rival labs respond on text detection.