Who Is Legally Liable When AI Kills Or Injures? Part 1

13/04/2018

The last few weeks have not been good for proponents of the superior safety of Artificial Intelligence (AI) over humans.  The Cambridge Analytica/Facebook scandal showed us that algorithms can and have been designed to analyse vast amounts of scraped data we freely provide the likes of Google and social media platforms, to manipulate the democratic process.  And a self-driving Uber vehicle became the first AI car to kill a pedestrian.  Elaine Herzberg was struck in Tempe, Arizona when the Uber car, which had a human inside to take over should an emergency occur, failed to slow down as she crossed the street with her bike.  She died in hospital of her injuries.

Of course, the Uber incident raises a variety of questions concerning the safety of self-driving cars.  However, there is another issue which arises from this tragic incident, namely, who or what is legally liable should AI kill or maim? 

Law is slow to catch up with technology

If there is one thing the Cambridge Analytica/Facebook data breach has taught us, it is the law is slow to catch up with technology.  At present, Big Tech companies operate with virtually no controls or regulations.  This week Mark Zuckerberg, founder and CEO of Facebook has been giving evidence before a hostile US Congress.  It seems almost certain now that the US will pass privacy laws to regulate internet companies.  

However the problem of damage occurring due to the actions of robots and the associated issue of liability is generally considered the most pressing questions by researchers, manufacturers, and other stakeholders[1]. 

Legal liability and AI

Dr John Kingston, senior lecturer at the University of Brighton has considered this question in a recent article entitled: Artificial Intelligence and Legal Liability[2].  Dr Kingston suggests that when it comes to self-driving vehicles, the laws that are likely to apply are those that deal with products with a faulty design.  However, he states, following such a path could hold back the development of self-driving vehicles as settlements for product design cases in the USA far exceed those involving human negligence, without including the costs associated with recalling the product.  Dr Kingston then goes on to argue that liability should, therefore, be dealt with under the laws of negligence, quoting Dobbs Law of Torts: 

“A bad state of mind is neither necessary nor sufficient to show negligence; conduct is everything”.

In his concluding comments, Dr Kingston states:

“There is also the question of who should be held liable. It will depend on which of Hallevy’s three models apply (perpetrator-by-another; natural-probable consequence; or direct liability):

• In a perpetrator-by-another offence, the person who instructs the AI system – either the user or the programmer – is likely to be found liable.          

• In a natural-or-probable-consequence offence, liability could fall on anyone who might have foreseen the product being used in the way it was; the programmer, the vendor (of a product), or the service provider. The user is less likely to be blamed unless the instructions that came with the product/service spell out the limitations of the system and the possible consequences of misuse in unusual detail.           

• AI programs may also be held liable for strict liability offences, in which case the programmer is likely to be found at fault.

However, in all cases where the programmer is deemed liable, there may be further debates whether the fault lies with the programmer; the program designer; the expert who provided the knowledge; or the manager who appointed the inadequate expert, program designer or programmer”.

Robolaw and the Guidelines on Regulating Robots

In 2012, the European Commission initiated a Robolaw Project, designed to investigate how emerging technologies such as AI should be regulated within the EU.  The outcome of this was a final report entitled: Guidelines on Regulating Robotics, designed to assist the European Commission create a solid framework for regulating bio-robotics[3].  The report states that scholars addressing this issue have clustered their responses into three groups:

1.    Limiting liability in a way so it does not preclude development of AI technology, excluding producers liability for “risks that could not be avoided notwithstanding the care in informing and designing the products”.

2.    Creating a legal ‘personhood’ for robots.  They would be then liable for the damage caused and, it would follow, be required to carry appropriate insurance.

3.    Impose strict liability on the owner of the AI, rather than the negligence standard because “the owner is a beneficiary of technology and can obtain additional advantages in introducing robots into his organization”.  It has been suggested that this method come with a ‘liability cap’ so as not to inhibit innovation.

The major concern with legal liability and AI is how to assign liability and regulate without hampering innovation.

In Part 2 of this article, we will examine how this might be achieved.