Anthropic to Watermark All Processed AI Content

AI firm Anthropic announced it will soon implement machine-readable watermarks on all content processed by its models, not just AI-generated output. This move aligns with the European Union's AI Act, which mandates that AI system providers watermark AI-generated or manipulated audio, image, text, and video outputs. The regulation applies to AI models released after August 2, with a grace period until December 2026 for existing models.

Anthropic confirmed that all new models offered globally will incorporate these watermarks from their launch. Text outputs will receive embedded, invisible watermarks, while other file types will feature digitally signed provenance metadata where supported. The company is adopting a broad approach, applying watermarks to all processed content even when the EU AI Act permits exemptions for assistive functions like grammar correction or minor edits that do not substantially alter meaning.

This comprehensive watermarking strategy means that model-level watermarks may tag content that the EU law intends to leave unregulated. The effectiveness and scope of these watermarks will become clearer once Anthropic releases a detection tool for public testing. The company stated its intention to eventually share details on detection methods to provide the technical support required by the EU legislation.