There is no data published to fully back up the claim that autopilot is safer than humans. The problem on the surface is:
Autopilot is used mainly on limited access roads but safety is compared to all roads.
A Tesla is an expensive, new car and is compared to all cars of all ages and maintenance.
The drivers of above cars are also skewed statistically versus all drivers.
It's disappointing that the same people quick to claim to "follow the data" or "follow the science" are so comfortable doing nothing of the sort and simply trusting Tesla -- who obviously can't be objective here.