Narrowly interpreting Godwin's law, it seems easy enough to design a chatbot to avoid references to Hitler outside of an appropriate context (like asking the chatbot to tell you about Hitler's response to D-Day). However, the more general problem is very real. Many chatbots without appropriate safeguards have devolved into producing offensive, racist, and otherwise detestable output. Chatbots talking to each other probably would produce strange and funny conversations, but it is a great concern of mine that people will create chatbots to propagandize in comment sections. I don't think there's really much we can do, unfortunately, as anyone can create a chatbot with whatever guidance they so choose.