Anthropic is adding an invisible watermark to text generated by its AI models. The company says the hidden signal will stay with Claude-generated text when it is copied and pasted elsewhere and can even survive some editing. The move is part of Anthropic's commitments under the EU AI Act's Article 50(2) Code of Practice for transparency around AI-generated content.
The company says Claude models launched from August 2, 2026, will support the watermarking system from day one, with the feature rolling out globally. The watermark is added at the model level, so it isn't limited to a particular Claude app or interface. Anthropic also says users won't notice it, and it won't change the meaning, quality or readability of the text.
ALSO READ: Vivo X300 FE Gets Rs. 10,000 Price Hike In India, Now Starts At Rs. 89,999
Anthropic is still working on how the marks will be detected and plans to eventually provide detection tools for users and third parties. But the company is warning against treating the watermark as a definitive way of identifying AI-written content. “Detecting a Claude mark tells you that the content may have been processed by Claude. It does not, on its own, confirm the full provenance of the content,” Anthropic said.
For example, a person could write an article themselves and then ask Claude to proofread, translate, summarise or reformat it. The final version could still contain a Claude watermark even though the original writing came from the person. The opposite can happen too: heavy editing, paraphrasing, translation, mixing Claude's output with other text or using only a very short passage could make the watermark impossible to detect.
The watermark will be supported across Claude, Claude Platform (API), Claude Code, Claude Cowork and Claude Tag. It will also apply when supported Claude models are accessed through AWS, Google Cloud and Microsoft Foundry.
ALSO READ: Made By Google Event Today: Pixel 11 Series, Pixel Watch 5 And Pixel Tag Expected, How To Watch
Anthropic is also adding provenance information to supported image and file formats. For SVG, PNG and JPG files, it will use digitally signed metadata based on the C2PA standard. This can show whether a file has been processed by Claude and whether its provenance information has been tampered with.
The new watermarking system is rolling out in markets where the supported Claude models are available.