Anthropic announced plans to add invisible machine-readable watermarks and provenance metadata to Claude outputs to comply with the EU AI Act transparency requirements starting August 2, 2026
Policy change ● Confirmed 78% confidence first seen
Anthropic said it will embed invisible, machine-readable labels in content produced by Claude models to help identify AI-generated text and comply with EU AI Act transparency commitments. The watermarking and signed provenance metadata are planned for Claude models released in the EU on or after August 2, 2026, with tools to detect and verify the marks and a later transition for previously released models.
Anthropic
Claude
August 2
C2PA
Claude Code
Claude Sonnet 5
Claude models
Code of Practice on Transparency of AI-Generated Content for Article 50 of the EU AI Act
December 2026
EU AI Act
EU AI Act Transparency Code
European Union AI Act
Reddit
Scarlet Letter watermark
invisible watermarks
Source coverage
Anthropic Plans Watermarks for AI-Generated Content from August 2026
Trending Topics
independent
Anthropic plans to add an invisible mark to AI text—as the industry scrambles to police AI slop
Fortune
independent
Some Claude users are mad that Anthropic’s new watermarks will catch them using it at their jobs, classes
TechCrunch
independent
Claude's new Scarlet Letter watermark is invisible — for now
Ars Technica
independent
How Claude's text watermarking works
Anthropic
official