Regulile Uniunii Europene obligă compania americană Anthropic să marcheze conținutul generat de chatbotul Claude

Anthropic plans to implement invisible watermarks in its AI-generated content to comply with EU regulations on AI transparency. This feature will be integrated into all Claude services, ensuring that generated files carry metadata indicating their AI origin. The watermarking techniques include imperceptible patterns in text and digital traces in files, which can be lost through certain editing processes. The initiative aligns with the European AI Act, which mandates marking synthetic content. However, the presence of a watermark only suggests AI involvement, not the complete history of the content.