Anthropic is implementing persistent watermarking across all Claude-generated content (text and files) to comply with the EU AI Act's transparency requirements. The watermarks use the C2PA open standard and are embedded at model level, meaning they survive copy-paste, export, and even minor text edits. All major AI platforms (OpenAI, Google, Meta, Microsoft, Black Forest Labs, Synthesia) have agreed to adopt similar compliance measures, signalling the end of undetectable AI use and the shift toward mandatory AI disclosure in content creation.