Artificial intelligence firms must reduce risks that include potential disruptions to services, communications and democracy, the UN's human right chief says.