I also find this mildly annoying, but I'm not entirely convinced it's any deliberate nerfing, it's acting a lot like a human would. If you asked a random person

I also find this mildly annoying, but I'm not entirely convinced it's any deliberate nerfing, it's acting a lot like a human would. If you asked a random person online "How do I take a duck home from the park?" you'd presumably get a lot of moralizing, and the internet is the dataset this was trained on.