Anthropic Will Watermark Claude’s AI-Generated Text Under New EU Rules

Anthropic says it will begin watermarking text generated by its AI models, including Claude, as part of its effort to comply with new European regulations. The company says models released after August 2 will automatically include technology that can identify AI-generated text and files.

The watermark will be embedded at the model level, meaning it will apply across Anthropic’s different products and surfaces. Anthropic also says the watermark can remain attached when text is copied and pasted and may survive some editing.

The move comes as AI companies increasingly face pressure from regulators and users to make AI-generated content identifiable.

Key Points

• Anthropic is watermarking AI-generated content: The company says all models released after August 2 will include watermarking for both generated text and files.
• The change is tied to EU regulations: The EU AI Act’s Transparency Code requires AI-generated or edited content to be marked in a way that other systems can identify.
• Older models will also receive support: Anthropic says it plans to extend watermarking to older models.
• The watermark travels with copied text: According to Anthropic, the watermark can remain when users copy and paste text elsewhere and may persist through some editing.
• It applies across Anthropic’s products: The company says watermarking will cover Claude, the Claude platform API, Claude Code, Claude Cowork, and Claude Tag.
• Files use C2PA: Anthropic says it is using the C2PA open standard for watermarking files.
• Questions remain about editing: Anthropic has not clarified how much editing would be required to remove the watermark.
• The broader industry is moving in the same direction: The article notes that Suno, Substack, Black Forest Labs, Google, Meta, Microsoft, OpenAI, and Synthesia have also taken steps toward identifying AI-generated content or adhering to the EU’s code.

Key Quotes

“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.”

Implications

Anthropic’s move signals that identifying AI-generated content is becoming a more significant part of how AI platforms operate, particularly as European transparency requirements take effect.

For users, the important detail is that Anthropic says the watermark will follow text beyond the original Claude interface. However, the article leaves an important question unanswered: how difficult will it be to remove the watermark through editing?

For the wider AI industry, Anthropic’s announcement also fits into a broader shift toward making AI-generated content identifiable, driven by both regulatory scrutiny and growing pressure around the use of AI-generated material.

Source: https://techcrunch.com/2026/08/11/anthropic-says-it-will-watermark-text-generated-by-its-ai-models/