Anthropic announced that Claude outputs will be marked with invisible, machine-readable watermarks and provenance metadata to comply with the EU AI Act transparency requirements
Feature update ● Confirmed 87% confidence first seen
Anthropic said it will embed invisible watermarks in Claude-generated text and add signed provenance metadata for supported image/vector outputs (using C2PA-style signals). The plan is tied to EU transparency requirements under the EU AI Act, with the watermarking expected to apply to Claude models released in the EU after the August 2, 2026 effective date and support for older models planned later.
Anthropic
Claude
Article 50
C2PA
Claude API
Claude Sonnet 5
EU
EU AI Act
EU Transparency Code
European Union AI Act
SiliconANGLE
The New Stack
The Register
Source coverage
Claude will apply invisible watermarks to AI text and images
The Verge
independent
Anthropic says it will watermark text generated by its AI models
TechCrunch
independent
Anthropic’s watermark survives copy-paste, but not the real dev workflow
The New Stack
Anthropic to start watermarking Claude-generated text, images
SiliconANGLE
independent