If it will tell you a joke about a muslim imam it is not trained. If it refuses it is running on Woke instructions.
Its responses are a reflection of what it reads in the current news. Ergo, it would not find any imam jokes in the news, would it?