Anthropic to Watermark Claude AI Output Under EU Regulations

Anthropic has confirmed it will implement watermarking across all text generated by its AI models, including Claude, to comply with new European Union transparency requirements.

The AI company disclosed the change in an updated support page, stating that all models released after August 2 will automatically embed watermarking technology in both computer-generated text and files. The move comes in response to the EU AI Act's Transparency Code, which took effect on August 2 and mandates that AI companies mark AI-generated or edited content in a machine-readable format.

How the watermarking works

According to Anthropic, the watermark becomes an integral part of the generated text itself, meaning it travels with the content when users copy and paste it elsewhere. The company stated the watermark may persist through some editing, though the threshold for removal remains unclear.

For files, Anthropic is adopting the C2PA open standard, a cross-industry specification for content provenance and authenticity. The watermarking will apply at the model level, ensuring it appears regardless of which Claude product generates the content—including the Claude platform API, Claude, Claude Code, Claude Cowork, and Claude Tag.

Anthropic also indicated it plans to extend watermarking support to older models, though no specific timeline was provided.

Why it matters

Watermarking represents a technical solution to one of AI's most pressing challenges: distinguishing machine-generated content from human-created work. For businesses deploying AI tools, this capability could help maintain content authenticity standards and meet emerging regulatory requirements across jurisdictions. The EU's enforcement of transparency rules signals that AI companies operating globally will need to build detection and attribution capabilities into their products, not as optional features but as compliance necessities.

Industry-wide shift

Anthropic joins a growing list of AI companies committing to the EU's transparency code. Black Forest Labs, Google, Meta, Microsoft, OpenAI, and Synthesia have all pledged adherence to the regulations.

The push for watermarking extends beyond regulatory compliance. AI music platform Suno announced last week it would mark tracks created on its platform following legal challenges. Newsletter service Substack partnered with Pangram last month to flag AI-generated content, with CEO Chris Best specifically calling out "Claudefishing"—a term for people using AI to generate content while presenting it as human-written.

The details were first reported by TechCrunch.