But look at your own argument. LLMs are not fair use because they might be prompted into regurgitating something substantially similar to the trained data.

And yet, the IA is 100% aiming to absolutely reproduce literally every part of the work in a 100% complete manner that replaces the original use of the work.

And you cannot bring yourself to admit that the IA is wrong. When you get to that point you have to admit to yourself that you're not making an argument your pushing a dogma.