AI company Anthropic recently announced that the new version of the Claude model, launched since August 2, will embed invisible watermarks in generated text to meet the transparency requirements of the EU's AI Act. These watermarks are imperceptible to the human eye, do not affect text readability, remain intact after copying and pasting, and may persist even after some editing. This move aims to prevent AI-generated content from being passed off as original and applies to content generated by Claude globally. Anthropic also plans to provide tools for detecting watermarks and is working to add this functionality to older model versions.