Anthropic to watermark Claude-generated text and files
Anthropic says it will watermark text and files generated by its AI models, including Claude, in response to the European Union’s AI Act Transparency Code. The company confirmed the change in an updated support page, stating that every model released after August 2 will automatically include technology to mark computer-generated text and files.
The requirement took effect on August 2 and calls on AI companies to mark AI-generated or AI-edited content in a form that other systems can identify. Anthropic said the measure will operate at the model level rather than being limited to a particular interface or customer product.
Watermarks across Claude products
That model-level approach means the watermark will be present regardless of where Claude output originates. Anthropic listed the Claude platform API, Claude, Claude Code, Claude Cowork and Claude Tag among the products to which the policy applies.
For text, Anthropic says the watermark is embedded in the output itself. It says the marker will travel when text is copied and pasted elsewhere and may persist through some editing. The company has not said how much editing would be required to remove the watermark.
For generated files, Anthropic is using the C2PA open standard. It also said it intends to extend watermarking support to older models, beyond those released after the August 2 date.
Transparency becomes an implementation issue
The move places content provenance alongside product deployment decisions for organisations using generative AI. Anthropic’s Claude tools are already being adapted for specialised professional workflows, including Claude for Legal plugins and integrations in legal work, where the origin and handling of generated material can matter alongside output quality.
Anthropic is not alone in committing to the EU code. Black Forest Labs, Google, Meta, Microsoft, OpenAI and Synthesia have also committed to follow it. The industry is moving toward marking generated content amid user criticism and regulatory attention.
Other platforms have recently announced related steps. AI music service Suno said it would mark tracks created on its platform after legal challenges, while newsletter platform Substack partnered with Pangram to flag AI-generated content. Substack chief executive Chris Best referred to “Claudefishing,” describing people who use AI to generate content.
What businesses should do
For businesses, Anthropic’s approach means AI-output controls should account for content provenance across APIs, workplace tools and downstream publishing channels. Teams should identify where Claude-generated text or files enter their processes and align review, disclosure and record-handling practices with the watermarking now built into the models.