INTERNATIONAL ARBITRATION IN THE AGE OF ARTIFICIAL INTELLIGENCE: OPPORTUNITIES AND EMERGING CHALLENGES

Artificial intelligence (hereinafter also “AI”) has become an integral part of modern society. While its use is sometimes criticized, and concerns remain about the potential replacement of human activity in certain fields, there is no denying that this emerging technology enhances work efficiency and simplifies many tasks. Given the rapid pace of development, those who choose not to engage with AI risk falling behind. The use of AI-based systems in the legal field offers numerous advantages, but it also presents significant challenges. The same applies in the context of international arbitration. In the sections that follow, the reader will find an overview of the main areas in which AI systems are currently being applied in arbitration proceedings (Section 1). This will be followed by an examination of the key challenges that arise from the use of such technologies in this context (Section 2). A brief overview of institutional rules and initiatives that have begun to address the integration of AI into arbitration will then be provided (Section 3). Finally, some concluding remarks and reflections on the way forward will be offered (Section 4).

1. Potential Applications of Artificial Intelligence in Arbitration

In international arbitration, AI can be highly valuable. It enables the automation of numerous legal tasks, helping to ease the workload of lawyers and arbitrators while enhancing the overall efficiency of proceedings. This technology can assist arbitration practitioners throughout the entire arbitral process, from the initiation of proceedings to their conclusion. Among its various applications, AI is particularly effective in analyzing large volumes of documents. It facilitates both document review and fact-finding and streamlines the document production phase. For example, AI can be used to quickly generate chronologies, accurately identify documents by category, and redact sensitive information more efficiently.^[1] AI can also accelerate and improve the accuracy of legal research, enabling practitioners to quickly access relevant case law and precedents. It can provide high-quality machine translations, which simplify communication in international arbitrations and help reduce translation costs for annexes. Moreover, the drafting of standard documents – such as procedural orders, procedural calendars, or Redfern schedules – can be efficiently managed using AI-based systems.^[2] AI can also assist in analyzing arbitrator profiles, helping parties identify the most suitable candidates based on objective and tailored criteria.^[3] In addition to the aspects already mentioned, AI systems also play a significant role in predictive analysis, offering parties and counsel advanced systems to preliminarily assess the likelihood of success in a given dispute or legal action.^[4] For example, several platforms available on the market use large case law databases to offer detailed insights into arbitral tribunals, judges, lawyers, and opposing parties. These systems claim to predict how different legal strategies may influence arbitral proceedings^[5] and their outcomes by analyzing established trends and complex statistical models.^[6] Some suggest that AI systems could even assist arbitrators in anticipating how an arbitral award might be treated in the future, such as the likelihood of it being set aside, refused enforcement, or freely enforced.^[7] This could help arbitrators draft decisions that, drawing on historical data and predictive models, are more likely to remain stable over time.^[8] Finally, AI could evolve to assist arbitrators in preparing the decision-making process, especially in highly complex disputes. These types of applications are already being used in ordinary litigation and class action proceedings.^[9] It is therefore possible that similar systems will also be adopted in international arbitration in the relatively near future (but see Section 2.4 below for potential challenges in this area).

2. Emerging challenges in the use of artificial intelligence

While the benefits of artificial intelligence outlined above are undeniable, it is essential not to overlook the potential risks linked to its use in international arbitration. The following section examines some of the main issues that may arise in this context. 2.1 Data processing and training: the “black box” and “bias” problem The “black box” problem in AI refers to the lack of transparency in how AI systems – particularly those based on complex algorithms like deep learning – reach their decisions.^[10] This lack of transparency makes it difficult to understand, interpret, and trust the outcomes generated by AI systems. In the legal field, it raises significant ethical concerns, particularly regarding the reproduction of discriminatory patterns, commonly referred to as bias. AI systems, after all, rely on the data they are fed.^[11] AI systems are trained on information that may include biased or unrepresentative data.^[12] If such data is not identified and corrected, the algorithm trained on it may perpetuate – or even amplify – certain biases, leading to inaccurate or flawed outcomes.^[13] The risk of perpetuating stereotypes may arise, for example, when AI systems are used to select arbitrators based on statistical probabilities drawn from available data, particularly data reflecting the prevailing composition of arbitral tribunals in practice.^[14] 2.2 Outdated data and “hallucinations” AI requires access to up-to-date and relevant data in order to function effectively. If the system is fed outdated information, there is a risk that its analyses and recommendations will be out of context or no longer applicable to current circumstances. In this regard, another significant challenge is the risk of so-called “hallucinations” generated by AI.^[15] These occur when AI systems generate inaccurate or misleading responses based on non-existent or incorrect data.^[16]  A notable example occurred during legal proceedings in the United States, where a defense team relied on an AI system to identify legal precedents to present at a hearing. It was only during the proceedings that the judge discovered those cases did not, in fact, exist.^[17] Another emblematic incident took place in early 2025 in Norway, where a lawyer submitted fictitious sources generated by an AI system during court proceedings. This prompted the Norwegian Supreme Court to update its guidelines, introducing a new provision specifically aimed at regulating the use of AI in judicial proceedings.^[18] These cases clearly highlighted the risks associated with the use of such technologies, particularly the tendency of some systems to produce inaccurate, misleading, or entirely fabricated content. In the context of arbitration, this raises the very real possibility that AI-based systems could lead to the submission of documents containing errors or flawed assessments, especially if the underlying data is poorly managed.^[19] It is therefore essential for lawyers and arbitrators to adopt a critical approach to AI technologies, ensuring through careful oversight that the information used is accurate, verifiable, and reliable.^[20] 2.3 Duty of professional secrecy, data confidentiality In addition to the points already mentioned, another important aspect must be considered, as lawyers are bound by a strict duty of professional secrecy.^[21] In an increasingly digitized age marked by the widespread use of AI systems, data relating to legal proceedings may not be fully protected when shared with AI solution providers or external platforms. In particular, there is a concrete risk that client information – such as case details or confidential documents – could be processed without adequate safeguards, potentially resulting in breaches of confidentiality and improper use.^[22] Online legal platforms and AI providers often do not clearly disclose whether, and to what extent, they may reuse retained data, raising concerns about the potential use of client information.^[23] In addition, legal data such as court decisions or contracts in full-text format are particularly difficult to anonymize effectively. Even when anonymization is promised or applied, contextual elements may still allow the identification of the clients or entities involved.^[24] Considering the above, it is essential for those who intend to use AI systems in arbitration proceedings – whether lawyers relying on them to draft pleadings or arbitrators using them for case management support – to carefully select their AI service providers. This helps ensure that confidentiality is respected and data is properly protected, avoiding any breach of professional secrecy.^[25] 2.4 Risks related to automated decision-making Finally, to complete this overview of critical issues surrounding the use of AI in arbitration, we turn to the risks linked to the potential use of AI by arbitrators in the decision-making process. As discussed above, this possibility may become increasingly common in the future. At the same time, it is essential to recognize that the role of the arbitrator must comply with specific ethical standards, including independence, impartiality, competence, and the duty to ensure the enforceability of awards.^[26] The improper use of AI-based systems by arbitrators to assist in their decision-making process could undermine these principles and may provide grounds for challenging the award.^[27] In this regard, a key consideration is the intuitu personae nature of the arbitral mandate, which requires that the assigned functions be performed directly by the arbitrator.^[28] Where delegation does occur – for example, through the appointment of a tribunal secretary – it must be disclosed to the parties in advance.^[29] Thus, legitimate concerns have been raised as to whether it is appropriate for an arbitrator to delegate activities related to the decision-making function to an AI system or software.^[30] Under French law, for instance, only a natural person with full legal capacity may act as an arbitrator.^[31]  Similarly, under Italian law, “no person who lacks, in whole or in part, the legal capacity to act may be an arbitrator”.^[32] In addition, case law in the context of AI-assisted patent registration has established that AI systems cannot be considered natural persons.^[33] Consequently, delegating adjudicative functions to an AI system could violate this requirement^[34] and potentially lead to a challenge of the award.^[35] In addition, it is worth noting that the New York Convention, which governs the enforcement of arbitral awards across many jurisdictions, provides under Article V that enforcement may be refused if the arbitration procedure was not conducted in accordance with the parties’ agreement or the law of the seat. Accordingly, the use of AI in the decision-making process, without party consent or in breach of applicable procedural law, could also risk refusal of enforcement under such Convention.^[36] A further issue concerns the alleged inability of AI systems – at least at this stage of technological development – to produce adequately reasoned legal decisions.^[37] Reasoning, which involves explaining the logical and legal grounds on which a decision is based, is a cornerstone of legal decision-making. This limitation represents a significant obstacle to the adoption of AI in arbitration.^[38] In light of these considerations, the use of AI systems in the decision-making process remains particularly problematic. It is thus essential that arbitrators and the parties involved agree on strict criteria governing the possible use of AI systems, to ensure compliance with the principles of independence, impartiality, and transparency throughout the arbitration process, thereby avoiding uncertainty and potential disputes. In this context, a key question emerges: should parties be informed when arbitrators or lawyers make use of AI during legal proceedings? This is an issue that will need to be addressed in the near future.^[39]

3. Emerging regulatory approaches on this issue

Precisely for the reasons outlined above, arbitral institutions are beginning to adapt to the introduction of artificial intelligence in arbitration proceedings. On 30 April 2024, the Silicon Valley Arbitration and Mediation Center (“SVAMC”) published the first edition of its “Guidelines on the Use of Artificial Intelligence in Arbitration” (“Guidelines”). This was the first example of an arbitration institution incorporating AI provisions into its rules.^[40] The Judicial Arbitration and Mediation Services (“JAMS”), one of the leading providers of alternative dispute resolution services, also began adapting to the emergence of AI systems in arbitration proceedings. On 23 April 2024 – around the same time as the SVAMC’s Guideline – it issued the “Rules Governing Disputes Involving Artificial Intelligence Systems”, a set of provisions specifically designed to regulate disputes arising from the use of AI systems.[41] In October 2024, the Arbitration Institute of the Stockholm Chamber of Commerce (“SCC”) issued a guide dedicated to the use of artificial intelligence in arbitration proceedings conducted under its rules. The document sets out guidelines aimed at ensuring that the adoption of AI technologies does not compromise the integrity and impartiality of the arbitral process.^[42] Despite the differences and specific features of the individual sources cited, some common principles emerge, reflecting the risks outlined above. First, human oversight remains central to all three institutions (see Section 2.2 above). Under no circumstances may final decisions be delegated to AI. Arbitrators are permitted to use such systems only for support activities, such as factual analysis or document management, while remaining responsible for verifying their accuracy and fully accountable for the results produced. Transparency is another shared pillar (see Section 2.4 above), although with some variation. The SCC requires arbitrators to disclose any use of AI, enabling parties to be informed and, if necessary, to object. The SVAMC adopts a more flexible approach, leaving the decision to the discretion of the arbitral tribunal, depending on the circumstances of the case. JAMS, on the other hand, has introduced specific protocols to ensure that AI-generated outcomes are accessible and subject to review by all parties. Finally, all three sets of rules place particular emphasis on data protection, confidentiality, and security (see Section 2.3 above). Each Institution requires that AI systems comply with the highest standards of data protection and anonymization, and mandates verification of data storage and management policies. It is therefore likely that, following these initial efforts, more arbitral institutions will soon begin developing frameworks to regulate the use of AI systems in the proceedings they administer, in order to remain aligned with technological developments.[43]

4. Conclusion and recommendations

In conclusion, artificial intelligence presents significant opportunities to enhance efficiency in international arbitration over the long term, creating new possibilities for optimizing time and resources. However, these innovations also introduce novel risks that require careful regulation and oversight by both arbitrators and legal practitioners. As a result, an increasing number of arbitral institutions are progressively developing guidelines to support the responsible use of AI. Looking ahead, collaboration among technology developers, arbitral institutions, and legal professionals will be essential to fully harness the potential of AI, helping to shape an arbitration process that is safer, more accessible, and better tailored to the needs of the parties involved. In the meantime, while this regulatory landscape continues to evolve, we offer the following recommendations for the safe and effective use of AI in arbitration: 1) prioritise AI systems with transparent processes and regularly assess the risks and critical issues associated with their use. 2) Ensure that AI systems are updated with current data and carefully review the outputs they generate. 3) Select AI technology providers with care, verifying that they comply with strict data protection standards and applicable regulations. 4) Discuss and agree on the use of AI with clients, parties and arbitrators and clearly define how it will be integrated into the proceedings. 5) When acting as an arbitrator, always retain full decision-making authority, using AI solely as a support tool and ensuring that its use is disclosed to and accepted by the parties.

---