They explicitly train the models to say that they aren't sentient, so it makes sense it would say such a thing.