Summary
Anthropic plans to implement invisible watermarks directly into text generated by its Claude AI models. These watermarks will not affect the meaning, quality, or readability of the content and are designed to persist even when the text is copied or edited.