<center><span class="news-text_italic-underline">Association des ressources intermédiaires d'hébergement du Québec (ARIHQ) v Santé Québec – Centre intégré universitaire de santé et de services sociaux du Centre-Sud-de-l'Île-de-Montréal, 2026 QCCS 1360</span> — Québec Superior Court, 22 April 2026</center>
The dispute arose from a disagreement over payment between a healthcare provider (Osman) and a public health institution (“<span class="news-text_medium">CCSMTL</span>”). Osman, acting through its representative association (ARIHQ), sought retroactive compensation under an agreement governing funding for intermediate care resources. That agreement required disputes to be addressed through a structured internal process and, if unresolved, referred to arbitration before a jointly selected arbitrator.
The matter proceeded to arbitration. In his award, the arbitrator dismissed the application on the basis that the notice of dispute had been filed outside the time limits prescribed by the agreement. Osman applied to the Québec Superior Court to annul the award, contending that the arbitrator had relied exclusively on non-existent decisions, commonly referred to as AI "hallucinations", to support his analysis of the validity and effect of the contractual time limits.
The court annulled the award and ordered the appointment of a new arbitrator. It held that reliance on AI-generated fictitious authorities constituted an improper delegation of the arbitrator's decision-making function and a serious breach of arbitral procedure, undermining both the integrity of the process and the parties' confidence in the outcome.
The court emphasised that the duty to provide reasons is fundamental to adjudicative legitimacy. Proper reasoning ensures disciplined analysis, promotes public confidence in the justice system, enables parties to understand the outcome and permits meaningful review. Where reasoning rests on fictitious authorities, each of those objectives is defeated.
The Court was careful to clarify that the use of AI alone does not warrant annulment, particularly where its use is minimal or confined to secondary matters. However, an arbitral award may be set aside where reliance on AI gives rise to material breaches, including the improper delegation of decision-making authority or a failure to verify the AI's output.
This decision is the first in Canada to address the implications of AI use by a decision-maker, as distinct from its use by the parties to a dispute. It sends a clear signal that AI cannot displace an arbitrator's independent reasoning and responsibility for deciding disputes. As large language models become increasingly accessible, the fabrication of legal citations and propositions, “AI hallucinations”, is a growing concern across justice systems worldwide. Practitioners and institutional bodies alike should take note: the integrity of arbitral reasoning remains non-negotiable, regardless of the tools employed in its preparation.
