I. Introduction.
Artificial intelligence (“AI”) may be described synthetically as “the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task ofusing computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable"[1] or simpler as “a branch of computer science dealing with the simulation of intelligent behavior in computers” or as “the capability of a machine to imitate intelligent human behavior"[2].
As described above, AI is nowadays one of the hottest topics in several areas. This reality takes place also in most Law areas, including Civil Liability Law. As relevant and current as it is, AI has not yet been specifically regulated in Uruguay in general terms nor has been regulated the impact of AI on Civil Liability.
In the following sections, the most important aspects of AI and Civil Liability will be considered in order to bring some first impressions of its connection.
II. Civil Liability in Uruguay. Brief description of the current regime.
Civil Liability is currently regulated under the Civil Code of Uruguay in almost the exact same terms as designed by the legislator in 1869. Specifically, most of the changes that took place in the Civil Liability legal regime over the last 150 years were not originated in a formal Law passed by the Legislative Branch but were originated in scholar and court opinions regarding the text of said Civil Code.
It is easy to acknowledge that over the last 150 years many aspects of life have changed substantially, and that civil liability as regulated in Uruguay must have been adapted to be applicable to new events not expected at the time of the passing of the Civil Code.
Among the changes that society have experienced in the last 150 years in Uruguay that needed to be considered under the Civil Liability rules, some situations that may be highlighted are the damages caused by automobiles, planes, hackers, computers, etc. It is clear that the creators of the Civil Code could not foresee each of the aspects mentioned above. However, the same Civil Code has been successfully applied to Civil Liability cases in Uruguay until today.
The Civil Code regulations on Civil Liability determine a dual regime of contractual liability and non-contractual liability. Contractual liability includes as per the opinion of the majority of courts and scholars any liability arising from the breach of a preexisting obligation, including obligations originated by sources different from contracts.
On the other hand, non-contractual liability include liability arising from the breach of the duty to not harm third parties (such as damages suffered by drivers in car accidents or in general damages caused to health, life or property).
In both regimes, the most accepted criterion is that, depending on the case, Civil Liability may be triggered by a strict liability regime or by a fault – based liability regime.
III. Civil Liability and AI in Uruguay.
Now, considering the definition set forth above, the first thing to consider is how that description shall be examined under the Civil Liability regime in Uruguay. Taking into consideration that machines using AI are not persons under the laws of Uruguay, normally, Civil Liability issues shall not be examined under the contractual liability regime, where the breach of an obligation shall be analyzed with respect to the person that was obligated to perform a specific obligation.
III.a. Contractual Liability.
As stated above, contractual liability does not determine specific problems considering machines or systems using AI, as the liability shall be always attributable to the person (debtor) of the obligationthat was breached.
Contractual liability as construed in Uruguay today determines that the debtor is liable for any breach of contract, including for breaches caused by negligence attributable to its employees (as stated in Section 1555 of the Civil Code) or defects or flaws of machines or systems used to perform the undertaken obligations. In this sense, if a machine or systemusing AI were to be held materially liable for the defective performance of the obligation undertaken by the debtor, under Uruguayan Law, the machine or system shall be considered only as a means used by the debtor and the final liability shall be imposed to the debtor.
III.b. Noncontractual liability.
If any machine using AI causes any kind of damage to a third party under a non-contractual basis, in order to determine who is liable for such damages, first it is necessary to determine which regime of liability applies.
Damages caused by machines are normally regulated by Section 1324 of the Civil Code (Responsabilidad por el hecho de las cosas).The legal regime determines that the custodian is liable for damages caused by a machine under his or her custody. The custodian is the one and only person with the actual power of control of the machine and may have such role the owner or any other person controlling the machine. The custodian’s liability is attributed on a fault-based liability, but where the fault of the guardian is legally presumed.
Considering that the Civil Code was passed in 1869 and that AI could not have been foreseen by the legislators, one again it becomes necessary to consider if machines using AI may be subjected to the same legal regime described above. In this case, the particularities are given by the peculiar characteristics of the AI, which allow machines to have a certain independent range of movements and behavior.
The most obvious way to regulate Civil Liability for damages caused by machines using AI is with the specific regime established by Section 1324 of the Civil Code which regulates damages caused by machines and things in general. In order to determine if such regime is applicable, the first thing to consider is if the machines using AI and generally all devices using AI are “things”. This first issue may be answered affirmatively, which allows to continue the analysis of Section 1324 and its applicability tomachines and devices using AI.
The second issue to consider is that if the AI has as one of its main objectives to make intelligent machines that may have certain independent behavior it may be difficult to identify a person acting as custodian, who shall be liable for all damages caused. However, at least in an elementary analysis, it may be plausible to identify one custodian who has the effective power of control of the machine or the AI system. In this scenario, the custodian may be the owner of the device, the manufacturer of the machine, the software provider or any other person that had the effective possibility tocontrol the machine or the AI system that caused damages. As the custody shall be individual, it may be hard to identify which was the failure that caused the machine to cause damages, but if that is not impossible, there are no strong issues that may prevent the application of Section 1324 of the Civil Code as with any other machine.
IV. Conclusions.
As briefly explained above, if the Legislative Branch does not approve new Laws creating a new civil liability regime for machines or devices using AI, Section 1324 of the Civil Code still may be applicable taking an evolutive interpretation stand point with this subject, such as happened with damages caused by automobiles, hackers, computers and other matters that were unexpected at the time of the passing of the Civil Code.
If you need any legal assistance regarding this or other related matters, please do not hesitate on asking.
[1] McCarthy, John, What isArtificial Intelligence?, Computer Science Department, Stanford University,in https://www-formal.stanford.edu/jmc/whatisai.pdf
[2]As defined by Webster Dictionary: https://www.merriam- zebster.com/dictionary/artificial%20intelligence
