It's the age-old problem of garbage in, garbage out combined with the inherent bias or error of whoever selects the training data.

It's like complaining that A.I. turns racist/sexist after it was "trained" on data including edgelord forums and 4chan.