Anthropic to Embed Invisible Watermarks at the Model Level for Claude

kimmonismus · x · 2026-08-11

Anthropic announced that invisible watermarks will be integrated directly into Claude models at the model level. Complying with EU regulations, Claude models launched in the EU on or after August 2, 2026, will embed machine-readable marks in generated text, applying globally across Claude APIs, Claude Code, and partner platforms.

Additionally, generated files like PNGs, JPGs, and SVGs will include signed C2PA provenance metadata where supported. The text watermark persists through copying and light editing but may be broken by heavy paraphrasing, translation, or in very short passages.

Related event: Claude Rolls Out Invisible Watermarks and Metadata, Sparking Compliance and Tech Debates(115 posts)→

Original post →

More from Models

Models channel →