The outputs stem from the inputs it was trained on, and the prompt that was given.

It's been trained on data to align the outputs to Elon's world view.

This isn't surprising.