In practice, sober automobile drivers who stay at the scene of a crash can kill people in an astoundingly negligent fashion without facing jail time: http://www.nytimes.com/2014/09/30/opinion/why-drivers-get-aw... This is a major reason I am excited about automated vehicles, along with the potential to unclog the madness that is urban parking politics, which is one of the major factors preventing the building of new neighborhoods that are truly urban.
In case of regular cars, you go to jail for vehicular manslaughter when you brake the traffic laws and in effect kill someone.
My guess is that, if the driver of a self-driving car decides to press the "auto drive" button and the car violates traffic laws and kills someone, it's still the driver who's responsible. The self-driving features are not deemed safe by law and are not meant to replace a driver.
It may change when self-driving cars are actually legalized as a thing (as it is currently discussed in California), but I can't imagine the current designs being close to passing a proper homologation process.
The answer for the moment is that we haven't figured out the ethical and legal questions in this area yet.
But the truck won't do that. Break the law, that is. Unlike humans, its knowledge of traffic laws will be complete. It won't "break the law" in the sense that you are using the phrase.
However, there are other ethical/moral issues (do we program a car to kill 1 person if it saves 3?).
Its knowledge of the text may be, its knowledge of what that means in practical terms will probably not be; the law often specifies that action be "reasonable" or similar terms, in light of the conditions. A human trier of fact may well find that the actions of an automated system -- especially in combinations of circumstances its creators did not specifically account for -- is unreasonable.