Anthropic Watermark Plan Makes AI Text Disclosure A Model Feature
Fast Company Middle East reported that Anthropic plans to add invisible text watermarks to future Claude models globally, aligning the rollout with EU transparency rules and provenance metadata for generated media.
Anthropic is preparing to make AI-generated text identifiable at the model layer, turning disclosure from a user-labeling choice into an embedded output feature.
Fast Company Middle East reported that the company described the plan in a support document tied to European Union transparency rules.
The rollout applies to future Claude models rather than only to European users.
Models released after August 2, 2026, will include machine-readable watermarks in text from launch, while existing models covered by the EU requirement have until December 2 to add support.
Compliance Moves Into The Output
The policy link is Article 50(2) of the EU AI Act's Code of Practice on Transparency of AI-Generated Content, which Anthropic has signed.
The watermark is intended to stay with text when it is copied and pasted, although forthcoming technical documentation has not yet detailed the detection method.
That design changes the compliance burden for customers using generated text in documents, messages or publishing workflows.
If the mark travels with copied text, disclosure becomes part of the content trail instead of depending only on a visible label added by the person or company using Claude.
Media Provenance Follows A Separate Path
Generated images, video and audio will receive standard provenance data rather than the same text watermark approach.
Anthropic plans to use the Coalition for Content Provenance and Authenticity standard for that media layer, a move tied to California's AI transparency law.
The compliance field is already crowded.
The European Commission's signatory list had about 190 companies and organizations by the end of July 2026, including Anthropic, OpenAI, Google, Microsoft, Amazon, Mistral and Cohere.
Meta and xAI were not listed among the signatories .
For enterprise users, the immediate operating question is how these marks will interact with internal archives, customer communications and tools that transform generated text.
Anthropic has not yet published the full technical documentation, so buyers still need to see how reliably the marker survives editing, formatting and platform transfers.
