Anthropic to embed AI watermarks in Claude models to meet EU AI Act standards
The move ensures generated text carries persistent identifiers, a step taken as major tech firms navigate the European Union’s new transparency requirements for artificial intelligence.
Anthropic has confirmed it will implement watermarking for text generated by its artificial intelligence models, including Claude, to comply with the European Union AI Act’s Transparency Code. The company outlined the measures in an updated support page, stating that the initiative is designed to mark AI-generated or edited content in a manner that other systems can identify.
The transparency code, which took effect on 2 August 2026, mandates that AI companies label their outputs. Anthropic stated that all models released after this date will automatically feature the watermarking technology for both computer-generated text and files. For file-based outputs, the company is utilising the C2PA open standard.
The watermarking is embedded at the model level, meaning it will be present regardless of which Claude product or interface the text originates from. This includes the Claude platform API, Claude, Claude Code, Claude Cowork, and Claude Tag. The company noted that because the watermark is part of the text itself, it will travel with the content when users copy and paste it elsewhere, and may persist through some forms of editing.
Anthropic also confirmed it will extend support for watermarking to older models, in addition to those released after the regulatory deadline. However, the precise extent of editing required to remove the watermark remains unclear. The company indicated it may persist through some editing, and TechCrunch has contacted Anthropic for further clarification on this point, with an update pending.
This decision aligns Anthropic with a growing cohort of technology firms responding to regulatory scrutiny and user backlash regarding AI-generated content. Other major players, including Black Forest Labs, Google, Meta, Microsoft, OpenAI, and Synthesia, have also committed to adhering to the EU’s code. The industry shift follows recent developments such as AI music platform Suno announcing it will mark tracks created on its platform following legal challenges, and newsletter service Substack partnering with Pangram to flag AI content.
