Helen Toner, who studies AI safety and policy, says if we need to take AI warnings seriously, and we can’t rely on the tech industry to fix the problem.