Claude AI models to add watermarks to texts and files

Anthropic announced that Claude AI models will begin adding watermarks to generated texts and files after the company signed the EU's AI regulatory act. Initially, only new models launched after August 2, 2026, will include this feature; older versions will receive it later, following a transition period. Companies can embed hidden statistical patterns into generated text—imperceptible changes in word choice and structure—that can later be detected by specialized tools as output from a specific AI model. This is one way to meet regulators' transparency requirements for AI-generated content.