Anthropic now embeds an imperceptible watermark in Claude text output and signed provenance metadata in generated files, across the consumer app, the API, Claude Code and its cloud partners. The EU AI Act only requires this for the European market, but Anthropic applied it globally, and the mark records that text passed through Claude rather than that Claude wrote it.
Read the full story: Anthropic watermarks every Claude output, worldwide →
Transcript
Anthropic now embeds an invisible watermark in everything Claude outputs. Not just images. Plain text. And not just in Europe, where the law requires it, but everywhere Claude runs. The trigger is the EU AI Act, whose transparency rules took effect on August second. It covers the European market. Anthropic covered the world, because maintaining two output paths across the app, the API, Claude Code and three cloud providers costs more than just marking everything. That is the Brussels Effect in real time. Now the part most coverage is getting wrong. The mark proves text passed through Claude. It does not prove Claude wrote it. Paste your own draft in for a grammar fix, and the output comes back marked. A school treating that as proof of cheating will be wrong about every student who used it as a proofreader. The technology is honest. The interpretation built on top of it is where this goes wrong.