Anthropic will embed invisible watermarks in text generated by new Claude models starting August 2, 2026. The change meets transparency rules under Article 50 of the EU AI Act. Anthropic has signed the related Code of Practice. New models will mark content at the model level, so the signal appears across the API, Claude apps, Claude Code, and cloud partners such as AWS, Google Cloud, and Microsoft Foundry. The marks apply worldwide, not only in Europe. Users will not see the watermark while reading. It stays in the text when copied and pasted and may survive some editing. Supported files will carry signed C2PA provenance metadata instead. Older models will receive marking support later during the transition period. Anthropic plans to publish detection tools. A detected mark does not prove Claude wrote the original content, since the model is often used to edit or summarize human work.
Anthropic will embed invisible watermarks in text generated by new Claude models starting August 2, 2026. The change meets transparency rules under Article 50 of the EU AI Act. Anthropic has signed the related Code of Practice. New models will mark content at the model level, so the signal appears across the API, Claude apps, Claude Code, and cloud partners such as AWS, Google Cloud, and Microsoft Foundry. The marks apply worldwide, not only in Europe. Users will not see the watermark while reading. It stays in the text when copied and pasted and may survive some editing. Supported files will carry signed C2PA provenance metadata instead. Older models will receive marking support later during the transition period. Anthropic plans to publish detection tools. A detected mark does not prove Claude wrote the original content, since the model is often used to edit or summarize human work.