Everything anyone could say about bad AI driving could be said about bad human drivers. Nevertheless, Waymo has not had a single fatal accident despite many millions of passenger miles and is safer than human drivers.
Even if driverless cars killed more people than humans they would see mass adoption eventually. However they are subject to farr higher scrutiny than human drivers and even so make fewer mistakes, avoid accidents more frequently and can't get drunk, tired, angry, or distracted.
For the same reason, I’d probably never buy a home robot with more capabilities then a vacuum cleaner.
https://www.wired.com/story/kia-web-vulnerability-vehicle-ha...
But even if they can theoretically be hacked, so far Waymos are still safer and more reliable than human drivers. The biggest danger someone has riding in one is someone destroying it for vindictive reasons.
Solving liability in traffic collisions is basically a solved problem through the courts, and at least in the UK, liability is assigned in law to the vendor (more accurately, there’s a list of who’s responsible for stuff, I’m not certain if it’s possible to assume legal responsibility without being the vendor).