Claude Watermark Debacle

I’m not sure that I even understand what all of the backlash is about with regard to the Claude watermarking issue. They pretty much said that any output that you even ask it to proofreed will have the watermark embedded in it so the watermark existing doesn’t even prove that it was used to generate the content … their own disclaimer basically says “this watermark just means the model touched it in some way not that the content was generated by the model”. My real concern with them adding watermarks to content is a bit different than most. Since the content is just a refactoring of previously read content copied from other original sources anyway the watermarking sort of just bugs me because it feels like anthropic is declaring some sort of ownership over content that was never theirs to begin with. To play devils advocate for a minute though. The tool creating a definitive “fingerprint” in the very text also means that it can be scanned for and inadvertantly advertises how the data was processed and gives a potential attacker more information than just the contained text about how they might begin a campaign against the end-user… maybe?