Ask HN: Is misaligned AI less dangerous than AI controlled by a bad actor