If I write a book, is there a legal way I can stop LLMs from reading it?
A book that is only intended for humans to read, maybe making each buyer to sign something? I'm not sure to be honest. Even if it's just for a limited period of time, 50 years, 20 years.
Recent court rules have affirmed that it is ok for models to read copyrighted material as part of training, as long as they pay for it.
That being said, some AI companies are offering methods to opt-out
First, the obvious: legal ways to stop anything have never succeeded in stopping anything.
Second, what would you loose from your book being used to train LLMs and what would you gain from preventing that? Are you afraid of loost sales? Or maybe plagiarism? Those problems have existed for a loooong time, LLMs or not. Or is it that you don't generally like the current AI craze and don't want to fuel it?
include the gamer word on every page and the whole book might get filtered out from the dataset
No. Even if you tried the AI companies don’t care. You don’t have the resources to enforce it, and slap on the wrist fines don’t bother big tech companies.
Don't show it to anybody.