Anthropic introduces invisible watermarks in AI-generated content to comply with EU transparency rules, affecting all Claude models globally. This move aims to distinguish AI-assisted from AI-authored content.