Even worse: the junk being deliberately fed to AIs is already at the stage where the results are useless BUT those results are blindly believed by many people BECAUSE they're generated by AI and therefore supposedly automatically correct!
Think Google's disastrous launch of their image generator which would under no condition generate Caucasian people because it had been fed exclusively black people because of the political ideology if its creators.
I've noticed similar things happening with inputs of climate models while working for our national weather agency, inputs which were deliberately filtered to exclude historical temperature extremes because the people in charge wanted a specific, politically motivated, output. And that wasn't even AI yet, but the results of those models are used as input for AI to make more sweeping predictions!
Now imagine similar things happening to AI being used to perform medical diagnosis, or worse yet medical procedures, just to give one example.