Anthropic has announced that it will watermark text generated by its AI models, including Claude, in order to align with European regulations. The company confirmed this move through an updated support page. The EU AI Act’s Transparency Code, which went into effect on August 2, mandates that AI companies label AI-generated or modified content in a way that other systems can recognize.

According to Anthropic, all models released after August 2 will automatically include technology to watermark both computer-generated text and files. For files, the company is utilizing the C2PA open standard. The company also stated that it will extend watermarking support to older models, noting that the watermark will stay attached to the text even when users copy and paste it. “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 explains.

It remains unclear how much editing would be required to remove the watermark. Anthropic has been asked for clarification, and any updates will be shared if a response is received. The company noted that watermarking will apply to various products, such as the Claude platform API, Claude, Claude Code, Claude Cowork, and Claude Tag.

Platforms are increasingly moving to watermark AI-generated content in response to user backlash and to avoid regulatory scrutiny. Last week, AI music platform Suno announced it would mark tracks created on its platform following a series of legal challenges. The previous month, newsletter service Substack partnered with Pangram to flag AI-generated content, with CEO Chris Best highlighting the issue of “Claudefishing,” a term for people using AI to produce content. Beyond Anthropic, other companies including Black Forest Labs, Google, Meta, Microsoft, OpenAI, and Synthesia have committed to following the EU’s code.