Anthropic Embeds Hidden Watermarks in AI-Generated Text

Anthropic, an AI company that's set to go public soon, has developed a way to embed 'imperceptible' machine-readable watermarks into text generated by its new Claude models. This means that when users copy and paste AI-written content, the watermark will travel with it, making it easier to identify where the material originated.

The technology is part of Anthropic's commitment to the European Union's AI Act, which requires generative-AI providers to make synthetic output identifiable in a machine-readable format. The goal is to give users clearer signals about content provenance.

The watermark applies globally and works across various platforms, including Claude, Claude Code, Claude Cowork, Claude Tag, and Anthropic's API, as well as through AWS, Google Cloud, and Microsoft Foundry. Older models will also be retrofitted with the technology.

This could have significant implications for schools and publishers, which are grappling with how to handle undisclosed machine-written content. The watermark could provide an additional tool for detecting AI-generated work in educational settings or publishing environments.