Even if the system is safer than human drivers (I can’t speak to this), why would the lowest paid workers at the company voluntarily sign up for _any_ increased liability? If one of the programmers pushes out a bug that kills someone, why should an employee who had nothing to do with this be charged with a crime?

You raise a good point - if the system is so safe, the people who make such claims should accept some liability - but those are not the same people as the ones who sometimes intervene in the driving.