Anthropic will add invisible watermarks to text and files generated by Claude, complying with a newly enacted European regulation. The announcement triggered backlash, with some users reportedly canceling Claude subscriptions. The author argues text watermarking is a natural extension of labeling efforts for AI-generated images and videos, citing a Pew finding that only 4% of US adults use AI chatbots constantly and Slack data showing many desk workers hide AI use. Limitations of provenance tech like SynthID are noted.
No score is assigned. Sources and their independence are shown in the citation chain below.