AI chatbots have turned out to be full of flaws. One big problem: They don't understand the complicated human factor. Nevertheless, they are very convincing in conversations — which could put users in harm's way. Should consumer AI be more strictly regulated?