I think it would be safe to say, better than the average driver at a bare minimum
Absolutely not. You can’t really tell anything from this statement because none of their numbers are qualified.
Autopilot is only engagable by experienced drivers (not really, but vehicle cost roughly correlates to driving experience). Comparing with just Tesla’s removes this bias and shows a 1.7x improvement, not 6x.
Autopilot is typically only engaged on the safest roads. Drivers are expected to disengage in less than ideal conditions. This might account for most of the 1.7x improvement, it’s impossible to tell from the data given. For all we know these ultra safe miles where autopilot is typically engaged might be 1.8x safer than typical miles, with autopilot making them slightly more dangerous.
Autopilot is meant to supplement the human driver, not replace it. So “humans are safer drivers when using autopilot” might be true, but “autopilot is safer than human drivers” certainly isn’t.