Anthropic adds invisible watermarks to Claude AI-generated text

Anthropic is a company that builds smart computer programs called AI, and one of its programs is named Claude.

Claude can write stories, answer questions, and make pictures and files for people.

Starting August 2, everything Claude writes will get a secret hidden mark called a watermark.

You cannot see this mark when you read the text, but computers can find it, like a tiny invisible stamp that says a robot made this.

Anthropic is doing this because the European Union, a big group of countries, made a law called the AI Act.

The law says companies must be honest about when a computer creates something.

The hidden marks will work all over the world, not just in Europe.

Files made by Claude will also carry a digital note that shows where they came from and whether they were changed.

This can help teachers, writers, and readers know if a computer wrote something.

But rewriting the text a lot or taking a screenshot of a picture can make the hidden marks disappear.