Anthropic is watermarking text generated by Claude to comply with EU law

“Anthropic, the company behind the AI model Claude, has started watermarking text generated by Claude to comply with the European Union's new AI transparency rules. This means that text generated by Claude will have a hidden pattern that can only be decoded by someone with the key. The watermarking method does not affect the quality of Claude's output or slow it down, but it has limitations, such as not being able to tell whether Claude actually wrote the text or just edited it. Expert Analysis: This development highlights the growing importance of AI transparency and accountability in the EU, and may set a precedent for other countries to follow. As AI-generated content becomes more prevalent, watermarking could become a standard practice to ensure that users are aware of the source of the content. However, the limitations of the current method may lead to further research and development in this area.”