Because it is a legal question, not a technological one.
Now, I don’t know if the for example US traffic law has a tickbox somewhere a manufacturer can go and mark that they will take full responsibility in case of any accident and it will never be the result or liability of the owner/“driver” of the car, but until it does exist there is only supervised self driving, no matter how well or poorly it actually functions or what it does.
Even the current robotaxi endeavours are just one major fatal accident away from grinding to a halt when the courts start figuring out who in the chain from insurer to owner to manufacturer and every worker and designer who has even remotely touched the project is actually responsible for that death.
It is, machine learning, neural networks and all the other parts in LLMs and generative algorithms like midjourney etc are all fields of artificial intelligence. The AI Effect just means the goalposts for what people think of as “proper” AI are constantly moving.