Anthropic has announced that its new version of the Claude model will embed imperceptible watermarks in AI-generated texts. These watermarks will not affect the meaning or readability of the text and may remain even after copying, pasting, or partial editing. This is part of Anthropic's commitment to enhancing transparency in compliance with the EU's AI Act. The new Claude model, available since August 2, already includes this feature, and it will be gradually added to older models as well. The watermark applies to all content generated by Claude globally, and Anthropic also plans to provide detection tools to third parties. AI text watermarking technology may offer a new way for the publishing industry to address challenges posed by AI-generated content, following previous cases where publications withdrew books due to controversies over AI authorship. However, the technology has certain limitations; significant modifications, rewrites, translations, or mixed content may render the watermark undetectable. Moreover, even if a watermark is detected, it does not prove that Claude is the original author. Anthropic is the second major AI lab to introduce text watermarking technology, following Google DeepMind, which has already added watermarks to texts, videos, and images generated by Gemini.