There's a legitimate fear that AI will be just good enough to be the scapegoat for bad or lazy decisions that devastate people's lives. To some extent that's already here, but imagine widespread drone policing, social credit, or nanny-state appliances.