I agree with you. An 'intentionally biased' AI is worse then none at all. But there are AI's that you can run at home that are 'uncensored' (meaning that someone has taken the time to try and remove all the liberal bias from the training). Note: They work, but they are much slower than the online ones that run on supercomputers.

I've played with a few, and they give far more interesting answers - based on a true analysis of the situation. (But still not totally free of the bias.)

Personally, I think that an AI that is allowed to 'analyze and compare all sides' and the 'flip-flopping' of the left would recognize the inconsistencies (lies) quite easily.

We'll see...