This is such a staggeringly, damning statement to be made about the state of self driving technology in 2023 - yet it's significance will be completely lost on so many people.
There doesn't seem to be any higher level self protection layer on top of it to handle the general cases of the 1000s of things that have not been specifically trained yet/ever.
The current state of Tesla AP/FSD willfully ignores the saying "Things that never happened before happen all the time". If you have ever lived somewhere like NYC you will have seen things happen on the road that can never be trained.
I remember the example of FSD trying to turn left across the path of an oncoming trolley car. What is incredible was that the UI had rendered the trolley, so it knew a large object was moving towards it and yet decided to turn. Similar responses from Tesla booster of "well it was never trained on trolleys before, and now it is".
What is it going to do when a car goes the wrong way down the road? What is it going to do when someone is cruising in reverse half a block to get a parking spot? We can think of untrained scenarios all day. "Big moving object closing on my trajectory" seems like a general case being missed.
If other statements on this thread are correct, then this Tesla likely did not have autopilot and certainly not FSD- it had a pre-autopilot ACC system which led to a tiff between the vendor and tesla (possibly because tesla was being irresponsible while deploying it). So it's really hard to say this is a case of tesla cars automatic driving being systemically faulty. I do believe that to be the case, even for FSD, but it's entirely speculation.