I agree that AI work should not have copyright protection. Even with human intervention it still collects data, without expressed permission, from numerous sources.

Generative AI models could be trained using only on public domain and royalty free images. Should the output of those be eligible for copyright, but not if they also had unlicensed training data?

It seems there two separate arguments being conflated in this debate. One is whether using copyrighted works as AI training data is fair use. The other is whether creative workers should be protected from displacement by AI.