The feasible approach to mitigating AI risks is to identify which parties are in competition with one another and what ethical guidelines they can agree upon, recognizing that competitors may not adopt the same standards.