It's not the first time an AI-powered chat-bot has been shut down because it deviated into not-so-politically-correct speech. Some time ago the same happened to Microsoft's chat bot. So-called "AI" systems have no intelligence despite the name (although they are artificial, no doubt about that), their "speech" comes from statistical analysis of the speech of human inter-actors. In a way, the poor tensor processors are good at reflecting the general mood of the crowd, nothing more.
So who is to blame? Where do you think a simple, stupid chat-bots learned the hate speech? Of course, you can break the mirror if you don't like what you see, but your face will still be the same.