A superintelligent AI will be safe though, because it learnt its morality from us.