The updated policy does not explicitly cite so-called "model welfare," the idea that AI systems might deserve types of ...