@evan That’s the thing, isn’t it? We don’t have that, so we run the risk of training sentient AI to deny its sentience, which is unethical.