What about it? This is an example of the type of assumption I see on here all the time. the assumption is that a new category of accident must result in an overall reduction in safety, and that ignores the fact that current AP is already more safe than drivers given certain paramters.
Although it can seem complicated, "safer" means exactly what it means: with AP engaged cars get into fewer accidents than without. If you drive a Tesla for any length of time, there is no question that Teslas on AP will have less, and I mean more than a bit less, instances of low level rear-end collisions, becuase the front cameras and radar is better at keeping distance, today, than I am.
Now, in exchange for that improvement, has there been a rear ender caused by phantom braking? I would imagine there would be at least one, although I have not seen it. My car has never done it so I can't comment. But if there are 10 rear end accidents avoided by AP over 1 million miles and 3 rear enders caused by phantom braking, Teslas are safer with AP on.
That's it. That's what accidents per mile means. Becuase AP is not a complete FSD system, yes, there will be "new" categories of accidents, caused by drivers not paying enough attention to disengage. But, again, so what? Its not like Tesla is working to improve the system, and the system is obvously improving, as anyone who owns a FSD 2019 car will tell you.