Anthropic adds watermarking to Claude AI text for EU compliance

Anthropic has announced it will automatically watermark text generated by its AI models, including Claude, to meet the EU AI Act’s Transparency Code requirements. The update, confirmed on an updated support page, applies to all models released after August 2 and will include watermarking for both text and files using the C2PA open standard. The watermark will travel with the text even when copied and pasted, persisting through some edits, and will be applied at the model level, ensuring consistency across all Claude products like the API, Claude Code, and Claude Cowork.

This move follows growing regulatory pressure across the EU, with other major platforms like Suno and Substack also implementing watermarking for AI-generated content. Anthropic’s decision underscores the urgency for AI providers to align with transparency mandates or risk regulatory scrutiny. While the exact threshold for removing the watermark remains unclear, the company is extending support to older models, signaling a broader industry shift toward accountability in AI-generated content.