Anthropic has published an announcement detailing the operational mechanisms behind Claude’s text watermarking feature. This explanation provides insights into how the AI model embeds subtle, traceable marks within its generated text to ensure authenticity and origin identification.
Source: Anthropic