Then you don't get to argue by them being safer overall. That argument then amounts to lie when the actual topic is general driving ability.

Second, then use only data about human drivers in perfect situations to compare. Their accident rates also go down.