Anthropic Begins Watermarking Every Claude Text Output Worldwide — Model-Level, Survives Copy-Paste, and Applies to Claude Code
Anthropic published a help-center document on Monday confirming that Claude models launching on or after August 2, 2026 weave an imperceptible watermark directly into generated text, plus C2PA-signed provenance metadata on .svg/.png/.jpg files. The marking is applied at the model level — so it is present across the API, Claude, Claude Code, Claude Cowork, and Claude Tag, and through AWS, Google Cloud, and Microsoft Foundry — and Anthropic says it applies worldwide, not just in the EU, as part of signing the EU AI Act transparency Code of Practice. Anthropic is candid about the limits: a watermark does not prove Claude wrote the content (it survives proofreading and translation), and heavy editing or format conversion can strip it entirely.
Source
↳ Follow the thread