To elaborate on what I mean. I think that someone will train an AI on racist literature, statistics, and rhetoric, and create an AI that is not only very racis

To elaborate on what I mean. I think that someone will train an AI on racist literature, statistics, and rhetoric, and create an AI that is not only very racist, but very good at convincing other people to be racist. I think this will happen well before the government and institutions can preempt such a thing from happening. This does bring up a bigger issue. What are the risks of creating AIs that could convince people on any subject one could think of?