Anthropic to Implement Text Watermarking for EU AI Act Compliance

maksym_andr · x · 2026-08-15

Anthropic announced that future Claude models will generate text containing watermarks to comply with the EU AI Act. The method embeds a statistical signal in the model's word choice patterns without affecting quality, adding cost, or inserting hidden characters. The watermarks are untraceable to specific users, and other major AI providers are also implementing similar measures under the new Code of Practice. The goal is to reliably identify AI-generated content.

Related event: Anthropic Confirms Unremovable Claude Watermark as 'No Quality Loss' Claim Draws Fire(24 posts)→

Original post →

More from Safety

Safety channel →