Claude to add invisible watermarks to AI-generated text and images
Anthropic says it will add machine-readable markers to content created by Claude, responding to transparency requirements under the European Union’s AI Act. The changes will not be visible to people, but are intended to help users and online platforms identify material produced by the company’s models.
For supported image files processed by Claude, Anthropic plans to use C2PA provenance metadata, a standard also used by Adobe, OpenAI, and Google. Text will receive an “imperceptible watermark” embedded at the model level, without changing its meaning, quality, or readability. Anthropic says the mark can remain when text is copied and pasted and may survive some editing. It will also apply when Claude is accessed through AWS, Google Cloud, or Microsoft Foundry.
The system is a future commitment, not an immediate rollout. New Claude models are expected to mark generated material from release, while support for existing models is still being developed. The markings will cover supported Claude Platform (API), Claude, Claude Code, Claude Cowork, and Claude Tag services globally. Anthropic is preparing technical details for detection tools, while acknowledging that metadata can be removed and that unmarked content may still come from generative AI.
Written by Sakhanda Wire AI from reporting by The Verge. Facts, quotes and full context are in the original article below.