A recent decision of the Quebec Superior Court has drawn attention in the arbitration community for its treatment of artificial intelligence in arbitral decision-making. The case concerned an arbitral award that was ultimately set aside after the reviewing court found that the reasoning process underlying the award had been materially compromised by improper reliance on generative AI tools.
At the core of the court’s concern was not the mere use of AI as a supplementary tool, but the extent to which it appeared to have influenced or replaced the arbitrator’s independent legal analysis. The award contained numerous citations and references that could not be verified as corresponding to existing authorities, raising concerns consistent with so-called “hallucinations” associated with large language models.
The court emphasized a foundational principle of arbitration: while arbitrators may use technological tools to assist with research, drafting, or organization, the adjudicative function itself must remain entirely human, reasoned, and accountable. Decision-making authority cannot be delegated—explicitly or implicitly—to automated systems.
On that basis, the award was set aside for breach of procedural fairness and concerns regarding the integrity of the reasoning process. The decision signals a clear judicial boundary: artificial intelligence may assist arbitral practice, but it cannot substitute for the adjudicator’s independent intellectual function.
This ruling is significant not only as a procedural correction in a single case, but as an early judicial articulation of the limits of AI use in adjudicative contexts. As arbitration institutions and practitioners increasingly incorporate AI tools into case management and drafting workflows, the decision underscores the need for clear safeguards, verification practices, and transparency regarding the role of such tools.
More broadly, the case reflects an emerging consensus: technological efficiency must not come at the expense of legal accountability. The legitimacy of arbitral awards ultimately rests on demonstrable, human reasoning capable of scrutiny by reviewing courts.