Anthropic is implementing watermarking on all Claude model outputs (text and files) to comply with EU AI Act Article 50(2), effective for new models after August 2, 2026. Claude will embed imperceptible watermarks in text and C2PA metadata standards in files, aligning with Google and OpenAI's existing C2PA support. This enables future verification of AI-generated content and reflects growing regulatory and industry standardization around AI provenance tracking.