Anthropic announced it will watermark content processed by any of its models, not just generated content, to comply with the EU's AI Act. The machine-readable watermarks will be invisible in text and will include digital provenance metadata for other files.
Applied globally to all new models from day one, the watermarking goes beyond EU requirements by covering even minor assistive edits like grammar corrections. Providers have until December 2026 to update previously released models.
Comments