Anthropic, a company that builds artificial intelligence (AI) systems, has added a new watermarking feature to its programme called Claude. This watermark is invisible and is built into the AI’s design. It now works with all new versions of Claude, and older ones will get it soon.

Anthropic started this because the European Union (EU) has made a rule that AI-generated text and images must be easy to recognise as artificial. The EU wants AI companies to clearly show when content comes from a machine, especially for people and businesses in Europe.

Because of this rule, Anthropic is putting watermarking on Claude everywhere, not just in the EU. Now, all text made by Claude has a hidden mark, so special tools can check if the AI helped create it. This helps people know if something was written by a human or by AI.