Reading: Anthropic Claude Ai Watermarks set for August 2 under EU AI Act

Anthropic Claude Ai Watermarks set for August 2 under EU AI Act

Published
3 min read
Advertisement

Anthropic said it will start adding machine-readable watermarks to text generated by new Claude models on August 2, 2026, a change meant to make AI output easier to identify. The marks will be invisible to readers, but they are designed to stay with the text when it is copied and pasted, and some of them should still be detectable after editing.

The move matters now because it ties a dated product change to Article 50 of the EU AI Act. Anthropic has also signed the European Union’s Code of Practice covering AI-generated content, and it says the watermarking will apply worldwide, not only in Europe. That means users of Claude, Claude Code, Claude Cowork and Claude Tag should see marked output across the service, while cloud customers accessing supported Claude models through AWS, Google Cloud and Microsoft Foundry will get the same treatment.

For files, Anthropic said the signal will be different. Supported output can carry signed provenance information based on the C2PA standard, so a valid record can show that Claude processed a file and whether someone altered its provenance data. But the company also cautioned that neither signal should be treated as definitive proof of authorship. Claude can process material originally created by people, and it can be used to translate, summarize or edit existing work, which makes the boundary between human and machine output harder to pin down.

- Advertisement -

That caution is the part users are likely to test first. Anthropic said heavy editing can weaken a text watermark, short passages may not provide enough material for reliable detection, and ordinary processing can strip provenance information from files. In other words, the mark is meant to help identify generated content, but its absence will not prove that a human wrote the text.

Older Claude models will not get the same treatment right away. Anthropic said it is working to add marking capabilities during the EU AI Act’s transition period, and it plans to publish technical details for detecting the watermarks. The immediate shift starts with new Claude models on August 2, and the open question is how quickly the rest of the Claude lineup will be brought under the same system.

Advertisement
Share This Article