Claude now puts an invisible watermark in the text it writes. Anthropic says it's doing this to comply with the EU AI Act, which from 2 August 2026 requires AI providers serving the EU to mark AI-generated content. It applies worldwide: in the Claude app, Claude Code, the API and through cloud providers.
The newest Claude models all carry it, and Anthropic is adding it to older models over the coming months.
By the end of this guide, you'll know how the watermark works, which popular tricks do nothing, and how to work with Claude so the words you hand in or publish are your own. That last part is the only reliable way to remove it.
