Key facts
- Anthropic is adding an imperceptible watermark to AI-generated text.
- The watermark will be applied to text generated by Anthropic's AI models, including Claude.
- This feature is intended to comply with the EU AI Act's transparency requirements.
- The watermark is applied at the model level.
- The watermark is designed to travel with the text.
- Significant editing may remove the watermark.
- The EU AI Act mandates transparency for AI-generated content.
Anthropic is introducing a new feature that embeds an imperceptible watermark into text generated by its artificial intelligence models, including its Claude chatbot. This development is a direct response to the transparency requirements outlined in the European Union's AI Act. The watermark is intended to be applied at the model level, meaning it is integrated into the AI's output generation process. Its design goal is to travel with the text, allowing for the identification of AI-generated content. However, Anthropic acknowledges that the watermark may be removable through significant editing or manipulation of the text. This initiative by Anthropic seeks to address the increasing challenge of distinguishing between human-created and AI-generated content, a key concern driving regulatory efforts like the EU AI Act. The act aims to foster trust and accountability in the rapidly evolving field of artificial intelligence by ensuring users are aware when they are interacting with or consuming AI-produced material. The implementation of watermarking is one of several technical approaches being explored globally to meet these new regulatory standards.