That's why I don't think 1 in 29000 is good enough given the amount of cars, people, and objects a car would typically encounter on regular basis. It could mean the disengagement is just to prevent it from crossing over to the next lane when no one is there, but it could also mean it was stopped from hitting an undetected pedestrian. I think an official risk classification needs to be developed, just like we have in commercial aviation, so we can tell just how safe the system really is.The report does require the reason for a disengagement such as:
Test Driver Street Disengage for unwanted maneuver of the vehicle that was undesirable under the circumstances
AV System Highway Disengage for adverse weather conditions experienced during testing
Without disabling the system when it's compromised in bad weather, the potential accident could be minor in slow, traffic jam speed, or major if the speed was at 90MPH
BTW, I know the reporting system for autonomous operation is one per x number of miles, so the typical airplane risk matrix would not work since those are 1 in x number of flights. So perhaps 1 in 1 million miles in autonomous car would be equivalent to 1e-9 in airplane?