Life exists for survival and propagation and organisms will change their immediate objectives to do so.

AI is built for a purpose. That purpose, however complex will always be a specific orientation because it exists and learns from its immediate stimuli - but once training stops, thats that. It must operate within the bounds of the conceptual box it was built from and can't diverge from it. No matter how well intentioned the goal is, there will always be boundary conditions in which the goal it was trained for will turn pathological.