The fact is that there is always a human in the loop. Without human supervision these algorithms deliver a small but significant portion of incredibly stupid results. So an actual human has to sit down, analyze these results one by one and decide what to do (in some cases just hardcoding the "correct" answer). The general public must be educated about this stuff so that responsibility is not muddled.
Incidentally, this also explains why there is zero interest in making rulebooks available, conscise, searchable, etc. All of these would improve fairness, but rulebooks are actually an instrument of power, not of fairness, so existing power structures will typically oppose any such changes.