Overview
- Anthropic said new Claude models launched in the EU on or after Aug. 2 will include an imperceptible, machine-readable watermark in text and C2PA-signed provenance metadata for supported files.
- The text watermark is applied at the model level so it travels with copied text and may survive some editing, while images and other files receive cryptographic provenance that can show Claude processed the asset.
- Anthropic warns the marks are not definitive proof of authorship because human-written work that Claude edits can be marked and heavy editing, short passages, file conversions, or older models can make marks undetectable.
- The company plans to retrofit older Claude models during the EU transition period and says it will publish tools and technical documentation for detection, but it has not released those methods yet.
- The announcement prompted swift user and developer pushback over proofreading, coding and attribution workflows and raises a broader trade-off between making marks robust enough to survive edits and keeping methods opaque to avoid easy circumvention.