Anthropic is implementing imperceptible, model-level watermarks directly into Claude's AI-generated text to comply with emerging regulatory frameworks like the European Union's Artificial Intelligence Act. The markers persist through light editing and copying.