I also doubt implementation of this system comes at any significant cost to counterterrorist efforts.

Your carjacking hypothetical goes more towards acceleration than speed. Most carjackings occur while a vehicle is stopped; accelerating beyond the speed at which a person can sprint (e.g., 10 mph) should be sufficient in most cases, and the system does not control acceleration.

And the circumstances in which speeding is beneficial are extremely infrequent. In order to determine the value of this approach, you need to balance the .1/million times in which it is beneficial to speed against the thousands of fatalities per year caused by speeding vehicles. My guess is that the lives saved will more than outweigh the loss of the opportunity to drive like Natasha McElhone in “Ronin”.

I suspect lots of people like to speed, and believe they can do so safely, and the prospect of automatic speed controls bothers them because it would mean no more speeding. In fact, implementation of the system might require some considerable fine-tuning of speed limits, not only by area but also by time of day; many roads are marked too low on the assumption that enforcement is difficult. But even if car speeds were set imperfectly low in most situations, the lives and fuel saved may very well outweigh the inconvenience.