On: Text AI watermarks will always be trivial to remove

[nonethewiser]

>I feel like this is going to end up being like cookie laws. It sounds good, I don't know how any one benefits from it.

Does it sound good?

Either anthropic does not tell anyone the signal and only they can verify the text. Or they share the signal and everyone can verify it, which means everyone can bypass it and its just an inconvenience.

[throwaway27448 → nonethewiser]

Are they not offering an API to perform this check? Would be pretty simple to use any open model, even a weak local one, to basically tweak the text until it doesn't flag.

[nradov → nonethewiser]

There are limits to what you can hide in text. If there is a signal then obviously independent cryptographers will be able to find it.