That seems really brittle. The baseline for every vehicle should be safety in every situation. (Sometimes that might mean: pull over to the side of the road and stop.) Only once every vehicle is proved safe, can certain combinations of vehicles negotiate compromises to safety in pursuit of other goals like speed.
(responding to your comment and the other comment about a possibility of attackers)
I agree that the baseline should be independent, autonomous, safe operation of each vehicle without any network/cooperation, but in time it's going to go beyond that. I have no doubt that it will happen.
Once independent operation is proven to be reasonably safe and dependable for everyday use, it will have achieved a certain degree of safety and efficiency - significantly better (at least in safety) than humans, but still prone to occasional accidents (regardless of who/what is at fault), especially when human drivers are still allowed to drive on the same roads.
At that point, people will demand even higher levels of safety (similar to the nines of high availability), and greater efficiency (i.e., speed). Unfortunately, with human drivers sharing the same road, those two demands are at odds with each other; greater efficiency inherently reduces safety. At some point though, the demand will be great enough that we'll have to adopt something better - something with a higher level of coordination between vehicles, and where only computer-controlled vehicles can participate.
It certainly won't be pushed or adopted universally in one step, and it'll likely never cover all road infrastructure. It will be adopted first on high throughput routes, much like you have private toll roads today. Instead of paying a toll (or perhaps in addition), your car will enter the roadway by joining the network. Without network compatibility, you'll be unable to use the road.
Once you've joined the network, however, your vehicle can literally drive at breakneck speeds, avoiding other traffic by incredibly small margins, and all with significantly higher safety than non-networked roads.
Of course having a single network controlling vehicles opens up a possibility for attacks, just as modern commercial aircraft can be vulnerable to attack. It will be an issue, but not one without reasonable mitigation strategies and best practices. The "network" in question may not even be centralized in nature, but rather a decentralized, local mesh network with safeguards in place for each vehicle to detect and respond appropriately to situations where the network seems to be guiding vehicles outside safe bounds. There are a lot of ways that it could be implemented, but it'll be important (at least to me) that it's not all controlled by one government/corporation/entity, but instead rolled out as a set of interoperable industry standards.
Why such confidence in the future adoption of this system that you stipulate will be both dangerous and inconvenient? This might work as a scifi plot device, but it's not predictive.
