Saying AI output is public domain is a misinterpretation of that case. The only court case related to the monkey selfies was by PETA, and the court ruled that the monkeys don't hold the copyright. The court didn't rule on who, if anyone, holds the copyright; whether Slater holds the copyright or it's public domain is undecided. The official opinion of the US Copyright Office is that generative AI outputs can be copyrighted by a person "to the extent that their contributions qualify as authorship".
Yes, copyright requires a human - in much the same way that an LLM requires a prompt. There's a continuum of author involvement that has to be evaluated on a case-by-case basis.
It's worth noting in your case that no determination of copyright was actually made by a court (well, almost: a court ruled that the monkey definitely didn't have copyright and that even if it did they wouldn't make PETA its guardian).
Some opinions were issued but those are non-binding and eventually the photographer gave up because litigating copyright is expensive.
