How can an AI be racist? It doesn't have a concept of what is wrong or good. It's just a model of things that a lot of people say. So the problem is not that the AI is racist, but the data that was feed into it; and that's a problem.