THERUNDWN_ World

Ars Technica

Technology

Anthropic Plans Invisible Watermarks for Claude

Anthropic announced plans to embed invisible markers into content handled by its systems to satisfy European Union regulations.

Added 8/14/2026

Pour mon site
Usertoop, CC0, via Wikimedia CommonsCC0Wikimedia Commons

According to Ars Technica, these machine-readable markings will apply globally to new models starting immediately, utilizing embedded text markers and digital provenance metadata for various file types. The company chose a broad implementation strategy that tags any processed material, including human-created writing that merely received minor grammatical corrections.

While the European Union rules offer exemptions for basic editing tasks that do not fundamentally change the original text, the current system cannot distinguish between complete generation and minor revisions. Future detection tools will be necessary to properly evaluate the effectiveness of the system, and these verification details are expected to be released later.

Share this story

Keep reading

All stories

What's next

Go back to feed

Were you satisfied with this article?

Original reporting at Ars Technica

STAY IN THE LOOP

Join the Rundwn.