The autonomous car and the dilemma of responsibility
We all love a good dilemma, the sort that brush aside our assumptions. On this occasion I’m not sure if this dilemma holds up, but it’s certainly an interesting conundrum over an scenario that seems pretty new: who is responsible when an autonomous vehicle has an accident and the software it uses is open source?
The question is raised by the development of an open source software, Openpilot, made available on Github by Comma.ai, the company created by George Hotz, who, at age 17 in August 2007, became the first hacker to unlock an iPhone so it would work with other operators.
The software is only valid at the moment for two specific vehicle models, the Acura ILX 2016 with AcuraWatch, and the Honda Civic 2016 Touring Edition, although adaptations will soon be available for other brands. Any user can download the software, install it in their vehicle and enjoy a certain degree of autonomous driving.
The software is not specifically created by Hotz, and has an information page in which it is clearly specified that it is an alpha product, that faults have been detected faults and that there may be others, and that any problems that may arise is the responsibility of the user. However, given the nature of the product and the notorious difficulty in carrying out reliability tests for autonomous vehicles, some say that such a disclaimer is not enough.
Under Spanish law, for example, a modified vehicle is authorized to drive on the roads only when it been approved or when the modifications installed have passed an inspection. In this case, a vehicle in which the software that has been installed at the factory has been modified to rewrite it with an open source would very possibly invalidate the insurance, and that responsibility rests with the user. However, the issue is not so simple when we talk about brands, as in the case of Volvo or Tesla, which have already clearly announced that liability in case of accident with their vehicles when they are in autonomous mode will be directly assumed by them if a design issue is the problem.
In the case of open source software, which can be modified and redistributed by anyone, the question becomes more complex. A large amount of open source software includes disclaimers or warnings that the software is provided “as is” without further consequences and that the responsibility for its installation and use rests with the user.
While it has become clear over time that open source software tends to have fewer errors and faster correction cycles due to the scrutiny offered by a large number of eyes, the problem arises when the possible consequence of a failure isn’t simply that the software does not work or loss of data, but a potentially fatal accident.
As autonomous vehicles become more common, the issue of responsibility becomes more complex. There are already brands that update the software of their vehicles over night and simply notify the user in the morning, which raises the possibility that a car owner might decide to go for a non-official update that offer certain advantages, particularly when our perception of safety increases significantly due to significant decrease in accidents involving autonomous vehicles.
If, instead of widespread private ownership of autonomous vehicles, they are instead part of fleets that offer a service to the end user, then this obviously offers a much higher degree of control, in addition to avoiding some of the unexpected problems that many pundits are already beginning to anticipate.
The dilemma will probably never actually arise, particularly considering the legislation in many countries, but it does serve to remind us that autonomous vehicles are with us already, and that they will need their software updating, meaning that we really do need to rethink many issues that until now, were pretty much a given.
(En español, aquí)
