By Dan Jerker B. Svantesson
There are countless news stories and scientific publications illustrating how artificial intelligence (AI) will change the world. As far as law is concerned, discussions largely center around how AI systems such as IBM’s Watson will cause disruption in the legal industry. However, little attention has been directed at how AI might prove beneficial for the field of private international law.
Private international law has always been a complex discipline, and its application in the online environment has been particularly challenging, with both jurisdictional overreach and jurisdictional gaps. Primarily, this is due to the fact that the near-global reach of a person’s online activities will so easily expose that person to the jurisdiction and laws of a large number of countries. Thus, online users ranging from individuals to the largest online companies are subject to unpredictable legal consequences when using the Internet. It also places stress on courts and regulators as jurisdictional claims frequently exceed relevant enforcement capabilities. Indeed, broad jurisdictional claims may force regulators to be selective in terms of targets to pursue, which will arguably undermine the rule of law principle that all are treated equally before the law. Despite intensive work by some of the world’s brightest legal minds, we are seemingly still far from solutions to these difficult situations.
This Post outlines the ways in which AI might help solve some of these challenges in private international law, as well as some issues that must be considered before getting to that stage. In doing so, the Post focuses explicitly on private international law as applied to Internet activities, though the general discussion is relevant to private international law more broadly.
What is AI?
The Council of Europe defines AI as “a set of sciences, theories and techniques whose purpose is to reproduce by a machine the cognitive abilities of a human being. Current developments aim, for instance, to be able to entrust a machine with complex tasks previously delegated to a human.”
The Council also notes the distinction between what has been termed “strong” AI, with the ability to “contextualize very different specialized problems completely independently,” and “weak” to “moderate” AI, with the ability to “perform extremely well in their field of training.”
“Strong” AI is generally beyond the reach of current technologies. Fortunately, none of the tasks outlined below that AI may perform for private international law require “strong” AI; thus, the required technology already exists. All that remains is to:
- construct suitable modeling of private international law;
- ensure access to the required data; and
- feed the modeling and data into an appropriate AI system.
While the idea sounds relatively simple on paper, clearly much work lies ahead.
Private International Law Modeling – Finding the “Correct” Point of Departure
For “weak” to “moderate” AI, the starting point for utilizing AI to help address the concerns expressed above is to create an appropriate model of private international law applicable to the Internet world. Like pioneering scholars such as Phan Minh Dung and Giovanni Sartor, I propose doing so based on modular argumentation:
“Modular argumentation offers itself as an appropriate platform for representing private international laws and different national laws as it allows knowledge to be split in separate modules. Moreover, it enables the different knowledge modules (which may represent legal systems or part of them) to be used by referring to each one of them specific issues. This is done by calling the relevant module and asking it to answer specific queries.”
The limited modeling that has been done of private international law has focused on assessing whether a given country can: (1) claim jurisdiction (and which domestic court is competent), and (2) apply its laws to a given scenario. This approach is natural given that this is also very much the focus of private international law as a discipline (in addition to matters of declining jurisdiction, as well as recognition and enforcement). Yet it is not the emphasis of the task discussed here.
The number of laws a person is expected to comply with when engaging in online activities is not static; rather it varies depending on a range of factors and is context-specific. Imagine, for example, that person X in Canada sends an email to a person in China, relating to the activities of a person in Romania. In this example, primarily three countries’ laws (Canada, China, and Romania) are relevant. Imagine further that the same person X also posts information on a US social media site on which she has “friends” in 100 different countries. In the context of this latter activity, person X is exposed to the laws of a vast number of countries due to the great reach of the posting. Thus, for any specific activity, we can speak of a “contextual legal system” consisting of the norms of all relevant states’ laws that the person in question is expected to abide by.
Knowledge of the number of applicable laws can be a useful diagnostics tool to highlight situations of “hyperregulation” in which:
- the complexity of a party’s contextual legal system amounts to an insurmountable obstacle to legal compliance; and
- the potential for legal enforcement of (at least parts of) the laws that make up the contextual legal system is more than a theoretical possibility.
However, knowledge of the number of applicable laws does little to provide legal predictability unless we also know the countries to which the acting person is exposed. The obvious difficulties include accessing the relevant laws, getting over language barriers, and identifying which states’ laws may claim to be part of the relevant contextual legal system in the first place. Here, predictability requires nothing less than an examination of all the laws of all the states in the world, including their respective private international law rules – perhaps humanly impossible, but just the sort of task at which AI thrives.
As a clarifying point, there is a distinct difference between the domestic legal systems and the contextual legal systems discussed here. While the former typically seek to avoid contradictory norms, each contextual legal system is made up of norms from multiple states’ legal systems – norms that typically are neither coordinated nor harmonized. Thus, unsurprisingly, a contextual legal system to which a person is exposed may contain clashing norms; that is, the norms of one state may demand performance of actions that the norms of another state forbid, or the norms of one state may outline duties that directly contradict rights provided for by the norms of another state.
Where norms within a contextual legal system conflict with others, an advanced AI system could be equipped to prioritize some norms over others based on agreed criteria. A range of traditional private international law tools, such as forum non conveniens, might be applicable in this context. Other methods, such as recent work on value-consistency, will also prove valuable. Taken one step further, legal concepts from other fields of law such as competition law may prove useful in addressing clashes of norms.
In any situation involving clashing norms, it is not just a matter of determining which country’s laws shall be applied to the situation at the expense of all other laws, nor is it a simple matter of assessing whether a certain country’s laws apply to the situation. An advanced AI system could analyze a contextual legal system with norms stemming from several legal systems, all applying at once. After all, we should not restrict ourselves to something as crude as assessing whether a given country’s laws apply to the situation, because not all laws of a country are relevant in any given situation.
Imagine that legal person Y from state A enters into a purchase contract with natural person Z from state B. If state B wants to apply its consumer protection laws to the situation, those laws of state B need to have a substantial connection to the matter and state B must have a legitimate interest in applying those consumer protection laws. However, if state B, based on the same set of facts, wants to apply its corporate governance laws to Y, the connection is weaker and the interest in doing so is less legitimate. To take this example to the extreme, imagine that based on the mentioned scenario, state B wants to apply its marriage laws to all employees of legal person Y; then both the connection and the interest is non-existent.
Thus, any assessment of whether state B’s laws shall apply hinges on what norms that state seeks to apply. It is the applicability of individual norms of a certain state, rather than all its laws in their entirety, that should be in focus. This increased granularity ought to be reflected in private international law rules, and their modeling alike.
At the most basic level, the assessment of claims of jurisdiction, as well as the questions of which norms apply and how clashes between applicable norms are addressed should be guided by the following three criteria that may be seen as the jurisprudential core of the concept of jurisdiction, namely:
- Substantial connection – is there a substantial connection between, on the one hand, the matter and, on the other hand, the given state seeking to exercise jurisdiction and the norms it seeks to apply;
- Legitimate interests – does the given state seeking to exercise jurisdiction and apply some of its norms have a legitimate interest in the matter; and
- Interests balancing – is the exercise of jurisdiction and application of those norms reasonable given the balance between the state’s legitimate interests and other interests?
The modeling must take all of this into account.
The Data Needed and How to Get It
Not least due to the successes of the Free Access to Law Movement and the work, and international outreach, of organizations such as the Australasian Legal Information Institute (AustLII), a significant number of states make their laws available online, although often in their local languages.
For those states that currently do not make their laws freely available online, the laws can and ought to be digitized. This is an important step not just for the AI system envisaged here. It is arguably a necessity for any state wishing to have credibility in claiming that its laws apply to the online conduct of foreigners; after all, laws that cannot be accessed arguably are less legitimate. Further, accessibility is a prerequisite for the international influence of any given legal system.
Language differences, as well as differences in the meaning of legal concepts, will no doubt place heavy demands on any AI system constructed to meet the goals envisaged in this post. However, although in the past language barriers were substantial hurdles for humans seeking to access foreign laws, AI systems provide a promising chance to get over these barriers, now or in the near future.
Furthermore, while in an ideal world the AI system would be fed all the laws (including both primary and secondary sources) of all the countries in the world, a useful system could be devised with a considerably more limited dataset. For certain purposes, it may be sufficient that the AI system is fed a subset of laws of particular importance. These subsets may be identified in numerous ways, including by reference to specific states, specific areas of law (e.g. consumer contracting, defamation, or copyright), or specific sectoral needs (e.g. the laws particularly affecting social media or the laws particularly affecting online news publishers). At least for these more limited systems, work ought to commence immediately.
The Envisaged AI System and Its Uses
Once the modeling discussed above is completed, and the AI system has been fed the required data, the system will be capable of:
- Identifying the norms from multiple legal systems that together make up the relevant contextual legal system for a given activity; and
- Reconciling – or at least balancing – those norms in a manner that makes for a coherent system even where individual norms clash.
The question then arises of the specific uses for such a system. In answering that question there is no need to be prescriptive. However, it is possible to point to at least four such uses:
- Predicting Legal Risks. An AI system such as that proposed above may help both natural and legal persons predict their legal risk exposure when engaging in online activities. In other words, it stands to benefit many different participants in the online ecosystem from the average Internet user to large companies such as Facebook, Google, Microsoft, and Apple.
- Assisting lawyers and legal education. The AI system envisaged here may help lawyers construct informed and evidence-based arguments by, for example, highlighting clashes within their clients’ respective contextual legal systems. Such an AI system could also be used for legal education purposes: both as a tool for constructing and analyzing real legal cases, and for constructing hypothetical scenarios to which students have to apply their legal knowledge.
- Law reform. An AI system such as that called for may further help law reform initiatives, of which there are many in the field of private international law (consider for example the work of numerous International Law Association Committees). The system could, for example, help identify instances of hyper regulation and propose formulations to minimize instances of conflicting laws.
- Court support. At the extreme end of the scale, the type of AI system discussed may even be utilized by courts either to inform the court, or indeed, to directly or indirectly decide disputes.
It goes without saying that the requirements placed on the envisaged AI system, such as transparency, explainability, consistency, accountability, and fairness, may vary depending on the use to which the system is put. Clearly, for example, such factors would be paramount for a system used by courts.
Any time one writes about the future, the line between optimism and naivety is a thin one. On the one hand, one shouldn’t downplay the challenges of embarking on the mission to construct an AI-driven system to help revolutionize private international law. On the other hand, however, it would be a missed opportunity not to start envisioning what may be achieved when the technology makes such a mission more feasible.
Sufficiently advanced AI technology exists and is developing further at great speed. If humans are about to entrust their lives to AI as the driver of cars, and if AI is believed to have the ability to disrupt the legal industry, AI should also be utilized to solve some private international law challenges.
Professor Svantesson is a Professor in the Faculty of Law at Bond University in Australia. He is also an Associate at the Swedish Law & Informatics Research Institute and a Visiting Professor at Masaryk University in the Czech Republic.