Anthropic Hides Invisible Watermarks in Claude's Writing

Anthropic now embeds an invisible watermark into text produced by Claude, its AI assistant. The hidden mark does not change the wording, quality, or readability of the output. It survives copy-paste and some editing, unlike labels attached to files. Anthropic is building tools for others to detect it. New models will carry the mark from launch; older models are still being updated.