A note on the "Fatal accident rate" topic:

A car which uses touch panels for control and electronics for simple things like doors, and relies only on a camera to do self-driving (no lidar), is bound to be unsafe in multiple ways. Drivers make more mistakes because they aren't provided physical switches and levers, they make mistakes because they're distracted by light from screens, in emergencies there are issues with unlocking doors and finally, when machine vision fails, there is no backup to tell of a Big Old Obstacle right ahead.

A note on the "Spying" topic:

Mozilla compared privacy policies. They didn't actually take apart cars, intercept their comms or dissect manufacturers' apps. I wish there existed a review of actual behaviours, as opposed to policies.

A car's ability to actually spy on the user depends on its tech, not just the policy. So a very simple car (not a Tesla, obviously) with a very agressive policy of "we collect everything" would be safer than a smartphone on wheels with a modest policy. You can't spy if you haven't got sensors.

As far as your points about the fatal accidents stuff goes - maybe Tesla shouldn't call it Full Self Driving if it can't fully self-drive without killing you, and maybe their cars shouldn't go 0-60 in 2 seconds by default, if that proves too difficult for inexperienced drivers to control.