Anthropic will apply invisible watermarks to text and files generated by its AI models, including Claude. TechCrunch reported on Aug. 11 that the step is intended to comply with transparency requirements under the European Union’s AI Act, which took effect on Aug. 2.

Anthropic said it updated its support page and will automatically apply watermarking technology to all models released after Aug. 2. For files, it uses C2PA, an open standard. It plans to extend the feature to older models.

The watermark is applied in a way that is embedded in the text itself. Anthropic said it moves along when copied and pasted elsewhere and can remain even after some editing. Because it is applied at the model level rather than the product level, the same watermark is attached to text generated across services such as the Claude platform API, Claude, Claude Code, Claude Cowork and Claude Tag.

How much editing is needed to remove the watermark is not yet clear.

Moves to attach identification marks to AI-generated material are spreading to other platforms. Music AI Suno said last week it would add labels to tracks made on its platform, and Substack began labeling AI-generated content last month in cooperation with Pangram.

Black Forest Labs, Google, Meta, Microsoft, OpenAI and Synthesia also said they would join efforts to comply with the EU transparency code.