Why everyone assumes that AI will be the first perfect software that won't ever malfunction? It's like getting together to see if we can build a supernova in a lab.

A complex AI capable of running human society would be a being completely beyond our comprehension already that we would definitely not know or be able to control.