Anthropic to Watermark All Claude Text Starting Aug 2, 2026

Anthropic is rolling out invisible watermarks on Claude text globally starting Aug 2, 2026, for EU AI Act compliance. Even proofreading/formatting triggers marking. Watermarks survive copy-paste and some editing; files get C2PA provenance. Detection tools available. Contrails AI CEO notes gaps and workarounds (heavy paraphrasing breaks it). This is a significant privacy and transparency change—users should be aware their content may be tagged.