Since their inception, self-driving cars have lacked one major thing, and that's a comprehensive legal framework to sort out matters of liability and regulation in the absence of a driver. Establishing what's allowed, who has to pay up when things go wrong, and what penalties may be involved and who is liable to them is something that the National Highway Traffic Safety Administration has been promising to do for a good while now, and it looks as though they've not only done just that, but the legal framework that they've drawn up seems to be mostly good news for fully automated vehicles of the sort that Google wants to put on the road in mass numbers.
The main talking points of the new set of policies revolve around making things as easy as possible for manufacturers and programmers of self-driving cars, within reason. For starters, there are, of course, a set of minimum standards and best practices outlined in the new laws pertaining to the construction, quality, and performance of self-driving cars. A self-driving car may statistically be safer than one with a human behind the wheel, but that is no guarantee, so these vehicles still need to be crash-safe. However, the guideline is at a federal level - this means that manufacturers can make a single fleet or type of self-driving vehicle and see it valid and welcomed in all fifty states, and shouldn't have to worry about individual states like California raining on their parade.
The various levels of automation are referred to as SAE levels, and the laws allow for "highly autonomous vehicles" all the way up to the level of full automation, albeit with a number of caveats for safety's sake. Lower level vehicles are generally thought of as those where the human driver holds more responsibility, and only the top level, level 5, entails full autonomy of the vehicle. The new policies are effective immediately, which means that automakers and tech firms finally have a definite guideline to work with in creating their vehicles, and production can now move forward on a meaningful scale. While big players like Google and Uber are already testing their robot fleets in public right of ways, the technology is still far from being perfect at the moment, and we can expect to watch the technology iterate up until its eventual public rollout, and even beyond.