Arguments are not executable code or the final work product of… whatever it is they're arguing about. You have to continue interacting with these people and their AIs, so the underlying question of whether they're trustworthy and not evil always exists.