Anthropic to Implement Invisible Watermarks on AI-Generated Content

Anthropic, a leading AI model provider, has announced plans to watermark all content processed by its models, regardless of location. This move is in compliance with the European Union's AI Act, which requires AI system providers to watermark AI-generated or manipulated content. The law applies to AI models released after August 2, with a grace period until December 2026 for providers to update previously released models. Anthropic's new watermarking approach will be applied to all new models offered globally, including those used for standard editing tasks such as grammar correction. The watermarks will be invisible to users and will include embedded metadata in text outputs and digitally signed provenance metadata in other generated files. This approach goes beyond the EU's requirements, which only mandate watermarks for cases where an AI system substantially alters the user's content. The implementation of these watermarks is expected to enhance transparency and accountability in AI-generated content, particularly in the context of the EU's AI Act.