Anthropic extends hidden watermarking of Claude text to older models
From 3 November 2026, Anthropic will apply statistical watermarking to older Claude models too. A dedicated tool is meant to indicate that a text may involve AI, but it cannot give certainty.