Anthropic implements Claude text watermarking to comply with EU AI Act

Watermarking adds no cost, tokens, or quality impact, and can't be traced to individuals.

Other major model developers signed the same EU Code of Practice and will also implement watermarking
Watermarked and un-watermarked text won't be distinguishable to readers
Nothing is added to the text and there are no hidden characters

Claude users get watermarked outputs with no visible change, added cost, or traceability to them.

Sources