Sounds rather flawed to me
At first sight, this all sounds fine, but there seems to be some flaws.
The idea that someone has to maintain their car (including software) properly is just extending the MOT process to include software and patching. Effectively, someone must be within the 'supported' range of software versions. As you can't guarantee everyone will update instantly, it can't just be latest.
However, if the car when running in autonomous mode has an accident for which it is to blame, the insurance of the 'driver' (I know he's not driving at the time, but has given the car permission to drive itself) will carry the cost. At first sight, this is reasonable, but is it really?
When you drive a car, it is under your control (supposedly) and is probably (in general) maintained or not by yourself, so if anything happens, it's because you made an error or failed to maintain it properly etc. So, your insurance pays. However, if your tyre blew out due to a manufacturing defect, the tyre manufacturer could well become liable if proven.
Now take a car in automatic mode. You're running software you have no control over. You've made sure it's up to date etc. and done all you can, but the functioning of the software isn't something you can reasonably test or validate. When that car has a crash for which it is found liable, why should your insurance company (effectively you) pay? If the software has a bug in it (equivalent of a manufacturing defect), why should they not pay?
Biting the hand that feeds IT
