Anthropic has started embedding invisible watermarks into everything its Claude chatbot produces, a move that has already sparked backlash from users who fear the markers will tank the value of AI-assisted work.
The policy shift responds to the European Union AI Act’s transparency code, which took effect Aug. 2. That framework aims to help people tell the difference between human and machine-generated content. Anthropic confirmed the watermarking applies automatically to text, images, and files across every version of Claude offered worldwide, including the API, Claude Code, Claude Cowork, and Claude Tag.
“When a supported Claude model generates text, it weaves an imperceptible watermark directly into the text itself,” the company wrote on a support page. The marker remains invisible and does not affect meaning, quality, or readability. It also survives copying and pasting, and may persist through light editing.
Teacher complaints about AI-fueled cheating have grown louder in recent years. Meanwhile, some Claude users took to Reddit and X to vent about the new policy. One Reddit commenter called it “a massive blunder” and predicted the mark would become “a scarlet letter” for anyone trying to sell AI-written content. Another user on X said the approach solved little because free, open-source models without watermarks remain widely available.
Anthropic acknowledged its detection system has limits. Heavily edited, paraphrased, translated, or very short passages may escape identification.
The European Commission secured commitments from nearly 200 companies by July 31. Signatories include OpenAI, Meta, Google, and Microsoft. Regulators framed the pledge as a way to build public trust and reduce deception.
The push comes as publishers and platforms grapple with a surge of AI-generated material passed off as human work. Last month, a debut novelist lost a $2 million book deal after his publisher suspected the manuscript had been written with AI. As watermarking rolls out across major platforms, the friction between transparency rules and commercial AI use seems likely to intensify.
