Car manufacturers are excited about the new options for autonomous driving. Rightly so – it would be the greatest revolution in this field in decades. Autonomously driving cars are not just good to take a nap on a boring ride. They enable entirely new business models, new forms of mobility, they combine the relaxed luxury of a train ride with the individual flexibility of the car, and the ongoing parallel evolution of the information layer on top of everything, the cars may even be turned into robots, doing the groceries all by themselves or picking up the kids without any parents involved.

However – we’re not there yet. Over the past two weeks, Tesla cars on autopilot in beta testing have produced two serious crashes, one of them deadly for the car owner. The incidents are not fully understood yet – and Tesla will have a high incentive to make it look like it’s not the fault of Tesla – but it seems rather certain that it was the fault of the autopilot misreading the exterior environment in both cases. Serious improvements will have to be made, at least at Tesla.

But that is not the only problem car manufacturers have with self-driving cars. IT-security remains a high concern as well. If hackers can attack the autopilot functionality, anything can happen! And the autopilot will be connected to the Internet or some kind of Internet at least. Even mass casualties are conceivable, if hackers attack the cars through central server functions, sending out a devastating order to all cars of one manufacturer at once. An interesting target for criminals and terrorists.

Solving the security problem is not trivial. Security demands are high, yet security technology is not mature enough in this particular field. Standard solutions like firewalls or cryptography didn’t even manage to solve their core concern – normal office IT-security – over the past decades. The sad normal in this world is that you have to live with being hacked and try to reduce the damage. Embedded computers in a car are an entirely novel kind of IT, security-wise unknown and uncharted in technology, systemic effects, resiliences, zoning options, tactics, assets and many other things. Standard IT-sec will not provide sufficient security in the car for another decade (if ever). And the tolerances regarding safety and performance are different as well. You may live with an occasionally hacked desktop PC. But can you live with am occasionally hacked car brake?

Car manufacturers and OEMs still have to understand and to shape the IT-security market. Or they may fall for cheap fixes, consuming time and money without providing any difference apart from difficult dependencies. Sadly, the market for such conventional and likely inefficient technologies is on the rise. Just recently, the firm Frazer-Nash Research announced a software to enable the hack-proof car, to be rolled out on the London black cabs soon – an alleged solution for self-driving cars in the future. But the technology is just another firewall with a little crypto-application bolted on top of it. Likely no more than a minor nuisance for serious attackers.

In sum, the self-driving car still has a long way to go. Some problems may be solvable incrementally over time, even in beta testing with actual people in it. Others will not be solved any time soon. The automotive industry will have to learn to separate the one type of problem from the other, and it will have to learn to understand how to understand those more tricky problems – a very common problem among all “old” technologies going digital these days.

Finally, interesting (and quite pragmatic) ethical issue appear on the horizon of the self-driving car. A question raised recently among philosophers is how a self-driving car should react when an accident with many casualties may occur and the only solution would be to kill the driver (say, by not driving into the crowd, but into a wall). Surveys were prepared on this questions, and almost everyone agreed that the car should kill the driver. At the same time, however, no one in the survey would buy such a car.