Ideally we would have AI that doesn’t intentionally (“intentionally”) favour either side.
Here’s a great discussion on the bias of ChatGPT, where you can see that the model lies by omission or has negative things to say about one side and only positive for the other.