Doesnt AI learning its morality from humans makes it unsafe, I mean just look at some cases, humans dont exactly always have good morals