Tiffany Comprés, founding partner at Pierson Ferdinand LLP and co-chair of international disputes discusses AI usage in international arbitration
As global commerce expands, resolving disputes through international arbitration efficiently and fairly has become more important than ever. The rising complexity and volume of cross-border disputes call for new approaches to streamline the procedure, manage costs, and ensure consistency, particularly as the volume of data implicated in each case skyrockets. Artificial intelligence (AI) is increasingly instrumental in meeting these demands, yet its integration in international arbitration raises fundamental questions about ethics, neutrality, and the unique qualities of human expertise.
THE CURRENT AND NEAR-FUTURE STATE OF TECHNOLOGY IN INTERNATIONAL ARBITRATION
Today, AI is streamlining arbitration by automating routine tasks, such as assisting in reviewing documents and preparing initial drafts of procedural orders and party submissions. Machine learning tools sort through vast volumes of evidence, assisting counsel in building cases more efficiently and allowing arbitrators to assess them more effectively. These tools also sort case law and quickly identify patterns across thousands of previous cases, helping in the analysis of legal issues. AI-driven research platforms and predictive analytics, although still imperfect, allow lawyers to focus more on complex strategy and judgment rather than manual, time-consuming tasks.
Natural language processing (NLP) tools enable faster and more accurate document analysis and translation, which can be invaluable for cross-border cases involving multilingual documents and intricate factual details. AI can make notes during hearings for use by arbitrators and counsel and suggest counter-arguments to help counsel hone their cases. It can even take the role of a particular arbitrator (based on an analysis of prior awards) for mock hearings to help attorneys prepare their arguments. Damage experts will soon use AI to determine the value of losses and explain at the evidentiary hearing how their AI works to support the valuation. Engineers engaged as experts will have their AI produce videos of construction and mechanical issues central to the case at hand. Third-party arbitration funders will use AI to cull cases with low probabilities of success.
AI is not only transforming how routine tasks are handled, but it is also poised to take the place of the junior associate at law firms. Although today this technology is still in the early stages – reminiscent of a goofy first year associate that makes mistakes and doesn’t understand the nuances that make a great legal argument – given the exponential improvement in the technology, in short order AI may very well become the indispensable junior associate, churning out research and first drafts. As technology that is expensive today becomes broadly available at a lower cost, small businesses and consumers will have more direct access to sophisticated legal tools.
Beyond these advances, blockchain technology is revolutionizing contract enforcement. Blockchain-enabled smart contract arbitration mechanisms, such as Kleros, can self-execute under specific conditions, reducing the need for human intervention and eliminating the need for traditional disputes altogether. These systems fully automate every step of the arbitration process (securing evidence, selecting the trier(s) of fact, etc.). For small, routine disputes, AI can serve as the adjudicator if the parties so agree. This is especially valuable for straightforward, lower-stakes disputes that can benefit from quick resolutions without sacrificing fairness. For example, eBay has been using an automated resolution system for years, purportedly settling up to 90% of claims without any human input. You might have thought this was the future, but it’s already happening.
THE RISKS
New technology always comes with unintended consequences. While AI can bring efficiency to many aspects of international arbitration, there are many open questions for the future.
THE DATA SET: CONTENTS AND SIZE
An AI system needs to train on vast amounts of data sets to function properly. In international commercial arbitration, some awards are published in an anonymized fashion, but most awards remain confidential. The entire set would ideally be anonymized and published to provide the largest possible dataset to train AI. This has been an increasing trend, but is not yet the standard. Further, in investor-state arbitration, the universe of cases is relatively small, so this is a foundational challenge. Training an arbitration AI on court judgments is not a viable alternative because national courts often have a domestic bias even in the interpretation of international law, which would taint the entire data set. In addition, the procedures in international arbitration are vastly different from domestic court procedures, and domestic court decisions would therefore be useless in this regard.
A related issue is bias. Any AI is only as good as the dataset on which it is trained. Any dataset will naturally include the prevailing societal biases and legal biases. AI can entrench and even exacerbate existing demographic and legal biases instead of freeing us from them. Remember that at this stage AI is really only a statistical machine, albeit a very advanced one. As a consequence, the output will reflect the input.
LACK OF TRANSPARENCY
Stemming from the concern about the data set is the issue of transparency, often referred to as the “black box” issue, i.e., the impossibility of directly explaining the results or predictions of a given AI. AI arrives at solutions powered by a rationale opaque to even the most experienced programmer. As a “mere” user, to learn exactly what data any particular AI is trained on, much less to understand the complex algorithms that make it spring to life, is inconceivable. This lack of transparency in a legal system predicated on fairness is troubling.
This further leads to a web of issues when AI is used in international arbitration, chief among them whether this complex mathematical/statistical calculation is something akin to reasoning. This is critical in order to determine whether an international arbitration award written by an AI would be enforceable under the applicable international treaties and domestic laws, which often require “reasoned” awards.
Less frighteningly existential questions, with a more immediate and practical impact include: What level of AI are parties comfortable permitting an arbitrator to use? While clients are often eager for their lawyers to use AI to reduce costs, that enthusiasm may not always extend to the decision maker. Based on the current rate of technological evolution, AI’s capabilities will likely outpace clients’ comfort with AI playing a key role in deciding their dispute. Currently, institutions at times explicitly and implicitly limit the use of AI to draft an award – for example, the Guidelines on the Use of Artificial Intelligence in Arbitration from the Silicon Valley Arbitration & Mediation Center specifically prohibit arbitrators from delegating their decision-making function to AI. Will this eventually change as disputes become even more increasingly complex and document intensive?
There is little debate that, given AI’s current capabilities, human judgment remains essential for high-stakes, complex cases that demand nuance and cultural understanding. In international arbitration, sensitivity to diverse legal systems and cultural contexts is crucial. Automated tools, though powerful, can inadvertently introduce biases or oversights, particularly in cross-border disputes where regional differences play a critical role. AI is highly effective at pattern recognition, but it lacks the human capacity for empathy, intuition, and the ability to weigh subtle social, cultural, and legal factors that may influence the outcome of a dispute.
For this reason, a hybrid approach – where AI assists with routine and preliminary tasks while human arbitrators maintain oversight – can offer the best of both worlds. AI tools can accelerate initial case assessments, manage document intake, prepare chronologies, and facilitate communication among parties, but ultimate decision-making should rest with human experts, at least for the time being.
SAFEGUARDING CLIENT TRUST IN A TECHNOLOGICAL LANDSCAPE
Lastly, with the adoption of AI in international arbitration comes a responsibility to safeguard data privacy, uphold client confidentiality, and navigate regulatory compliance across borders. AI systems are only as secure as the frameworks governing their use. Professionals who adopt these tools must remain vigilant in overseeing how client information is managed, stored, and analyzed, in compliance with regulations like the GDPR and CCPA. A lapse in data security can severely impact client trust, which is foundational to any legal relationship.
As the legal profession embraces AI, international arbitration stands at the forefront of innovation. Hybrid arbitration models, combining AI-powered administrative support and junior attorney-level assistance with human-led judgment, may become the norm, blending efficiency with the integrity needed for high-value, complex cases.
