AI Policy for All Risks