this seems to be about cars that aren't totally driverless - ones where the "driver" can override the robots. If they find, for instance, that the problems are all caused by drivers, and none by the cars, then they can adjust their premiums accordingly when fully driverless cars come in.
The insurers say drivers need to be able to prove that they're not at fault if the technology goes wrong.
I don't understand that bit. As ZM says, defendants don't normally have to prove their innocence. Either the insurers or the BBC must have got that wrong?