On April 22, 2026, in Association des ressources intermédiaires d’hébergement du Québec (ARIHQ) c. Santé Québec – Centre intégré universitaire de santé et de services sociaux du Centre-Sud-de-l'Île-de-Montréal, the Superior Court of Quebec (the Court) overturned an arbitral award due to the arbitrator’s excessive reliance on AI. According to the Court, the arbitrator based his reasoning on non-existent sources “hallucinated” by artificial intelligence (AI) and thus “delegated part of his decision-making power,” thereby relinquishing his authority.
The decision, which is one of the first public decisions on the subject of AI in Canada, reinforces key safeguards governing the exercise of arbitral powers and raises important questions about the use of AI in legal decision-making.
Background
Osman Medical (Osman), the co-plaintiff with the Association des ressources intermédiaires d’hébergement du Québec (Quebec Association of Intermediate Housing Resources) (ARIHQ), acts as an intermediary accommodation provider for the Centre intégré universitaire de santé et de services sociaux du Centre-Sud-de-l’Île-de-Montréal (Integrated University Health and Social Services Center for the Centre-South of the Island of Montreal) (CCSMTL).
A dispute arose in 2021, when Osman claimed C$1,225,000 as compensation for housing accommodations provided between 2019 and 2021.
The national agreement governing the contractual relationship between Osman and the CCSMTL includes a dispute-resolution mechanism that imposes a strict 90-day deadline for submitting any dispute to the designated representative. However, the ARIHQ sent the notice of dispute three years after the dispute arose. The CCSMTL then sought dismissal of the arbitration request for failure to meet the 90-day deadline.
Arbitral Award
The arbitrator heard the parties and issued an arbitral award granting the motion to dismiss and finding that the forfeiture clause was valid, as it strictly sets out the time limit within which claims must be brought.
Court Decision
The crux of the case before the Court relates to the reliance of the arbitrator on AI. In particular, the arbitral award referred to an article that could not be traced as well as to four non-existent judicial decisions that contained legal citations to other existing decisions. Finally, a fifth decision, although real, was cited inaccurately and bore no relation to the principles invoked. These references and citations, as pointed out by the Court in its decision, constituted the majority of the legal basis for the award.
In light of this, the Court concluded that the award was null and void. It found that the use of “non-existent” and “hallucinated” sources supported the conclusion that “the arbitrator delegated his authority and abdicated his role in reviewing the outcome.”
Drawing on several fundamental principles of arbitration, including respecting the parties’ autonomy and intention, the importance of adequate grounds, and the duty to preserve the integrity of the decision-making process, the Court reiterated that arbitrators may not delegate their duties to a third party, including a technological tool such as AI.
The Court also emphasized the seriousness of the arbitrator’s misconduct, which undermines the parties’ confidence in the award, as well as the public’s confidence in the arbitration process.
The Court nevertheless took a nuanced approach to the use of AI. It acknowledged the risks inherent in its use, including the generation of erroneous references, the lack of human judgment and privacy concerns without, however, strictly prohibiting its use.
The Court also clarified that this award being overturned “does not imply that every award that cites erroneous references or uses artificial intelligence as a drafting tool should suffer the same fate.” Any assessment would depend on the circumstances of each case and take into account certain factors, such as the nature and extent of the procedural violation, as well as its actual impact on the award rendered. Thus, a minor breach would not necessarily result in the award being declared null and void.
Conclusion
Given that AI is expected to play an increasingly important role in arbitration, this decision is particularly important. While some arbitration bodies have begun to provide a framework for its use, notably through the adoption of guidelines, much uncertainty remains. For more information on this topic, please see our Blakes Bulletin: From Algorithms to Awards: Artificial Intelligence in the Arbitration Toolkit. This case illustrates the role of the courts in adapting fundamental legal principles to the emergence of these new tools. But above all, as the Court points out, despite technological advances, the responsibility of arbitrators to make the ultimate decisions in the cases before them remains at the heart of the arbitral process.
For more information, please contact the authors or any other member of our Arbitration or Litigation & Dispute Resolution groups.
