Free daily briefing on global business news.
The AI company signed the E.U. AI Act's transparency code and says new Claude models will mark content from day one of release
NurPhoto / Getty Images
Anthropic is adding machine-readable marks to text and images generated by its Claude models, a step toward complying with E.U. rules requiring AI companies to label AI-produced content. The markings are imperceptible to the human eye but enable people and platforms to identify when content has been produced by Claude.
The company uses two methods. Text generated by supported Claude models carries an embedded watermark that Anthropic says travels with the content if users copy and paste it and may hold up through some degree of editing. For files such as .png, .jpg, and .svg images, Anthropic is applying signed provenance metadata using the C2PA open standard, which is also used by Adobe $ADBE, OpenAI, and Google $GOOGL, according to The Verge.
