Autocars are going to be held to a much higher standard because they
should be held to a higher standard.
On one end, autocars will be held to a higher standard because of cultural fear of incredibly rapid technological progression. That fear is entirely rational; the bigger the change, the more potential failures, the bigger the risk that there are some hidden faults that no one predicted. Big changes require more prediction than incremental changes, and more prediction = higher risk that you're wrong. Saying you're 95% sure of just one thing isn't scary, saying that the all of these new, undealt-with-before elements all each have 95% chance of being fine is a much scarier combination, and the chance that just one prediction goes wrong that fucks up the rest of the sequence of predictions, messing up the system as a whole, that chance gets higher.
On another end, a fatally incorrect judgment by a Google Car goes into the database of encountered events that they have. The database that other Google Cars learn from. Which means the error could potentially compound. And other Google Cars will make the same mistakes, in higher frequencies. So there has to be some high standard to make sure false positives don't set about a positive feedback loop in the way the Cars perceive database history.