Trapping AIs via utility indifference — AI Alignment Forum