Anthropic will watermark text and files generated by its Claude models to comply with European rules, the company confirmed in an updated support page first reported by TechCrunch. The EU AI Act’s Transparency Code, which took effect Aug. 2, requires AI companies to mark AI-generated or edited content so other systems can identify it.
All Anthropic models released after Aug. 2 will automatically watermark their output, and the company said it will extend support to older models. For files, it is using the C2PA open standard for signed provenance data. The text watermark sits inside the words themselves.
“Because the watermark is part of the text, it will travel with the text when it’s copied and pasted elsewhere, and may persist through some editing. Watermarking will be applied at the model level, which means it will be present no matter which Claude product or surface the text comes from,” the support page reads. That covers the Claude platform API, the Claude apps, Claude Code, Claude Cowork and Claude Tag.
How much editing it takes to remove the mark isn’t clear. TechCrunch said it had asked Anthropic to clarify, and the company hasn’t published detection details or a public detector.
A hard problem for text
Text is much harder to watermark than images, because there is little room to hide a signal without changing what a reader sees. Software engineer Sean Goedecke, writing on July 2 ahead of the EU deadline, argued in a post titled “Text AI watermarks will always be trivial to remove” that any scheme strong enough to detect reliably costs output quality, and that rewording or running text through another model can strip it. He pointed to Google’s SynthID as one partial solution and to what he described as Unicode tricks used by OpenAI and Anthropic.
Other companies are moving the same way. AI music platform Suno said the week before that it will mark tracks made on its service after a series of legal challenges, TechCrunch reported, and in July Substack teamed up with Pangram to flag AI-generated newsletters.
Black Forest Labs, Google, Meta, Microsoft, OpenAI and Synthesia have also committed to the EU code, according to TechCrunch.
WebProNews is an iEntry Publication