Most people already own and 'control' some automated systems, and we have a functioning legal framework for establshing liability. Some examples:
1. An automated sprinkler system. If your sprinkler system sprays outside of your property onto some passerby's extremely expensive suit, you as the owner of the system are liable. You are in control of the system, have set it up to spray beyond the bounds of your property, and have given the system instructions on when to start and stop spraying, usualy via a timer. You could argue that the system is deficient, or there is a manufacturing fault which prevents the system from operating as intended, i.e. spraying where you tell it to within commmonly acceptable limits. If the nozzle comes broken and it sprays 360 degress instead of 60 degrees, you can attempt to pass the liability for the incident on to the manufacturer, assuming you can convince people that you didn't know about the defect in advance, in which case you should have acted to mitigate it.
2. An automatic robot vacuum cleaner. I have one, and while it's not a common thing in households yet, the liability surrounding it is pretty clear. You can block of certain areas with a magnetic strip that it comes with (it's a neato). In summer it's really hot, and I like to leave my front door open to get a breeze going through the house, but the neato can just fit under the security gate. So I put the magnetic strip across the front door. If I forget to put it there, and one day it goes off to clean in the street outside and it cause an accident, I'm liable, not Neato (the company). It's functioning as intended. I accept this risk as part of owning the robot. If it decides to go AWOL and *cross* the magnetic strip, not performing as intended, and cause an accident in the street, then Neato (the company) is responsible, or at least I have a good case to make them responsible.
Now why wouldn't an autonomous vehicle fit into the same framework. If an autonomous vehicle causes an accident while not under autonomous control, it isi definitely the drivers responsoibility. If an accident happens while the vehicle is under autonomous control and functioning as intended, it's the passenger's responsibility, because they presumably gave the instructions to the car, however since the instructions to the vehicle would be very high level, as in 'take me to X destination' and the intended function of the vehicle is to get you to that destination safely, you would have to have issude an order like "take me to X latitude, Y longitude" which happens to be just the wrong side of a cliff. Even then, the vehicle *should* avoid situations like these. If the vehicle does not perfrom as intended, i.e. it collides with something, which it is not meant to do, then it is the manufacturer's fault, or rather, again you have a good chance of making it their responsibility assuming you can prove the car would act consistenly incorrectly in the same situation. As far as disclaiming liability is concerned, make the autonomous systems get the euivalent driver's licenses, i.e. to sell a driverless car, it must be certified as capable of driving autonomously, which could be made to mean: will always act to avoid a collision in all directions, will drive on a particular side of the road, will negotiate with nearby autonomous vehicles using a standard protocol, will obey traffic signals, etc...
The legalities are not that hard!