Radmila Chapuis: “The question is: at what point does the use of AI constitute a delegation of authority that is incompatible with the role of commercial arbitration?”
Radmila Chapuis, Lecturer in Law at EDHEC Business School, examines the growing role played by artificial intelligence in arbitration procedures, which are now widely used by businesses as an alternative means of dispute resolution.
To begin with, could you explain what commercial arbitration is and why this method of dispute resolution is so popular today?
Radmila Chapuis: Arbitration is a method of dispute resolution characterised by the involvement of one or more arbitrators (1). It is referred to as an out-of-court procedure, meaning that it takes place outside the courts. All parties must agree to submit to arbitration, either in advance of or following the emergence of a dispute. The appointed arbitrator then settles the dispute by issuing a binding decision, known as an arbitration decision or ruling. This has the same legal force as a judgment handed down by a judge in a state court.
Why is this procedure so popular (2)? Firstly, it offers the advantage of flexibility: whether it be the language, the timetable or even the format of the hearings, everything can be adapted to operational constraints. Secondly, arbitrators are experts chosen for their sector-specific know-how: they therefore have a thorough understanding of the parties’ field of activity. It also, of course, offers potential time savings and greater control compared to litigation.
However, arbitration is also frequently chosen to limit public exposure that may be detrimental to the parties. Its purpose is to prevent the disclosure of confidential information and to avoid potential harm to the parties’ reputation arising from court proceedings (3).
Finally, the last but by no means least advantage for commercial disputes is the cross-border enforcement of the ruling, which allows the decision to be effectively enforced outside the country in which it was made (4).
In practical terms, how does an arbitration procedure work?
Radmila Chapuis: Firstly, pursuing arbitration necessarily implies that the parties agree on this method of dispute resolution. They must have understood that they are irrevocably waiving their right to state courts. The key word is consent!
Secondly, taking the dispute to arbitration must be formalised in writing, designating the arbitrator(s) and setting out the applicable rules. Arbitration therefore derives from an arbitration clause if agreed upon before the dispute arises, and from a submission agreement if the parties have agreed to arbitration after the dispute has arisen. The parties may agree on a wide range of matters, including the applicable law, the language of the proceedings, and the seat or venue of arbitration. Yet, only natural persons may serve as arbitrators. Legal entities, however, may be entrusted with the administration of arbitral proceedings. This form of arbitration, commonly referred to as institutional arbitration, is administered, for example, by the International Chamber of Commerce in Paris (5).
Finally, the arbitral tribunal resolves the dispute either in accordance with the applicable rules of law or as amiable compositeur, that is, on the basis of equity. Once rendered, the arbitral award has res judicata effect: it is binding upon the parties and conclusively disposes of the dispute.
What limitations (legal, economic or in terms of fairness) do you see in the growing use of arbitration?
Radmila Chapuis: First and foremost, I would say: the cost! It is indeed higher than that of court proceedings. The issue of the balance of power should not be overlooked either, as some agreements may be unbalanced for the parties (for example, in B2C contracts, where one of the parties is a private individual). The grounds for challenging the decision are also limited.
Thus, an appeal is only possible if the parties have provided for it in the arbitration agreement. Such recourse is of limited relevance, as the very purpose of arbitration is to avoid recourse to state courts, which would otherwise constitute the only competent authority to hear an appeal.
As for an action for annulment, it may be brought only on limited grounds, namely serious procedural or substantive defects, such as a breach of the adversarial principle, irregularities in the constitution of the tribunal, or an award contrary to public policy, as reflected in the New York Convention on the Recognition and Enforcement of Foreign Arbitral Awards.
Arbitration is now expanding into new areas, particularly disputes relating to cryptocurrencies and blockchain. Why is this sector making such extensive use of arbitration, and what specific challenges does this present?
Radmila Chapuis: Indeed, in recent years there has been a rise in litigation relating to cryptocurrencies, particularly in the United States (6). This increasing resort to litigation can be attributed to the expansion of digital assets, including exchange platforms, smart contracts, and fundraising mechanisms. As the financial stakes have grown, so too has the incidence of disputes.
The widespread use of arbitration can be explained by several factors. Firstly, these players operate in an international environment: platforms are registered in one jurisdiction, servers are located in another, and investors are spread across several continents. Arbitration offers a neutral solution, particularly suited to this type of ecosystem. Secondly, the technical nature of these disputes makes it attractive to appoint arbitrators with specific expertise. Finally, the confidentiality of arbitration is a major asset for companies whose reputation and valuation rely heavily on market confidence.
However, such disputes give rise to unprecedented challenges. The first stems from the pseudonymity inherent in blockchain technologies: a digital address does not necessarily correspond to an identifiable legal or natural person, thereby complicating the identification of the parties. The second difficulty concerns the location of assets. Cryptocurrencies are not situated within any specific physical territory; rather, they exist as units of value on a decentralised network. This lack of territorial anchoring raises complex issues of jurisdiction and the enforcement of arbitral awards. Enforcement, in particular, remains a critical concern: even where an award has been rendered, it is still necessary to identify assets capable of attachment and to establish a sufficient connecting factor with a given jurisdiction.
Let’s turn to your main point: what role does artificial intelligence play in arbitration procedures today? How is it actually used, and what legal or institutional frameworks govern these practices?
Radmila Chapuis: Today, AI is primarily accepted and used as a support tool. Arbitrators are increasingly using it for document analysis, translation, drafting, and so on. It is true that it saves a significant amount of time in these areas (7). At the institutional level, several arbitral centres have issued guidelines governing the use of artificial intelligence, notably the Silicon Valley Arbitration & Mediation Center (SVAMC) (8) and the Stockholm Chamber of Commerce (SCC) (9). These documents emphasise key principles: transparency regarding the use of tools, protection of confidentiality, and the maintenance of effective human oversight in decision-making.
Legal doctrine and best practice also highlight requirements regarding data protection, cybersecurity and compliance with the GDPR, notably through the joint protocol of the International Council for Commercial Arbitration of the New York City Bar Association and the International Institute for Conflict Prevention & Resolution (10).
Finally, at European level, the AI Act introduces a regulatory framework that imposes specific obligations on providers of AI systems and clarifies the applicable liability regimes (11).
Is AI likely to transform the very nature of arbitration in terms of timeframes, costs, or even the role of the arbitrator? Where does the line currently lie between technological assistance and decision-making authority?
Radmila Chapuis: Artificial intelligence does not replace the arbitrator’s decision-making function. The arbitrator retains full authority to assess the evidence, exercise independent judgment, and assume responsibility for the award. That said, the degree of procedural transparency varies depending on the applicable institutional rules and the parties’ agreement.
At present, case law remains limited; however, one frequently cited decision is LaPaglia v. Valve Corp. (2025) (12). The case concerns an application for annulment, alleging, in particular, that the arbitrator improperly delegated part of their duties to an artificial intelligence system. No judgment has yet been made.
Personally, I believe that the fundamental question is not “should AI be banned in arbitration?”, but rather: At what point does the use of a tool amount to an impermissible delegation of the arbitrator’s mission, and what are the implications for the arbitrator’s impartiality and the confidentiality of the award? It is certain that, in the coming years, we will see these issues clarified.
Given the rapid pace of these developments, what safeguards need to be put in place? Could it be conceivable, for example, that the inappropriate use of AI could undermine an arbitral decision? Where, in your view, should the line be drawn?
Radmila Chapuis: What emerges from current legal scholarship and the first instances of litigation is, above all, the obligation incumbent upon any decision-maker—whether judge or arbitrator—to ensure transparency in their use of artificial intelligence. AI must remain an auxiliary tool and must not be used to substitute the arbitrator’s decision-making function. Improper or uncontrolled use carries a real risk of undermining the validity of the decision, as well as the confidence placed in it by the parties.
The question of neutrality is equally critical: how can it be ensured that AI produces neutral outputs? The fundamental principles of arbitration—namely the impartiality of the tribunal, equality of the parties, and respect for the adversarial process—must be strictly preserved. Finally, particular attention must be paid to confidentiality and data security. This entails compliance with essential safeguards, such as prohibiting the unauthorised uploading of case materials and ensuring that any AI system used complies with applicable regulatory frameworks, including the GDPR.
In the future, shall AI become a genuine arbitrator (thus moving beyond its role as a mere assistant), it will be important to put robust safeguards in place. One might assume, for example, that the parties involved will need to have consented to the use of AI, but also be aware of the scope of its role as an arbitrator.
References
(1) https://www.dictionnaire-juridique.com/definition/arbitrage.php
(3) https://theses.fr/2015AIXM1093
(4) Chapuis, Radmila. Noms de domaine et modes alternatifs de règlement de conflits. Thèse de doctorat en droit privé, Université Paris 2 Panthéon-Assas, 2021 - https://hal.science/tel-04039315
(5) https://www.icc-france.fr/nos-actions/resolution-des-litiges-commerciaux-internationaux/arbitrage/
(7) https://www.village-justice.com/articles/arbitrage-efficacite-accrue-confiance-fragilisee,55712.html
(8) https://svamc.org/svamc-publishes-guidelines-on-the-use-of-artificial-intelligence-in-arbitration/
