Anthropic to Embed Imperceptible Watermarks in Claude Text for EU Compliance

lilyraynyc · x · 2026-08-11

To comply with the EU AI Act, Anthropic will weave imperceptible watermarks into text generated by future Claude models and add digitally signed provenance metadata to files. The feature will apply globally across Claude platforms and applications, with plans to update older models. While the move may boost the appeal of open-weight models, some users are skeptical about the watermark's effectiveness and potential interference.

Related event: Anthropic embeds invisible watermarks in all Claude outputs, responding to EU AI Act(49 posts)→

Original post →

More from Models

Models channel →