On April 22, 2026, the Superior Court of Quebec issued a significant decision 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, overturning an arbitral award after finding that the arbitrator had relied extensively on artificial intelligence-generated legal authorities that did not exist.
The ruling is among the first publicly reported Canadian decisions addressing the use of AI in arbitration. It highlights the importance of preserving judicial and arbitral independence while raising broader questions about the appropriate role of artificial intelligence in legal decision-making.
Background
Association des ressources intermédiaires d’hébergement du Québec (ARIHQ), together with Osman Medical, brought a claim arising from accommodation services provided to the Centre intégré universitaire de santé et de services sociaux du Centre-Sud-de-l’Île-de-Montréal (CCSMTL).
Osman Medical sought approximately C$1.225 million for accommodation services provided between 2019 and 2021.
The parties’ contractual relationship was governed by a national agreement containing a dispute resolution procedure requiring claims to be submitted within 90 days. However, the notice of dispute was not delivered until approximately three years after the dispute arose.
The CCSMTL therefore sought dismissal of the arbitration on the basis that the contractual limitation period had expired.
The Arbitration Decision
After hearing the parties, the arbitrator dismissed the claim, concluding that the contractual forfeiture provision was valid and that the claim had been brought outside the prescribed 90-day period.
The Superior Court’s Findings
The central issue before the Superior Court was not the contractual interpretation itself, but the arbitrator’s reliance on artificial intelligence during the preparation of the award.
The Court found that the decision relied upon:
- An academic article that could not be located;
- Four judicial decisions that did not exist;
- A genuine court decision that was inaccurately cited and unrelated to the legal principles being discussed.
These AI-generated—or “hallucinated”—sources formed a substantial part of the legal reasoning supporting the award.
Because the legal authorities underpinning the decision were largely fictitious, the Court concluded that the integrity of the arbitral process had been compromised.
AI Cannot Replace an Arbitrator’s Judgment
The Court held that the arbitrator had effectively delegated part of the decision-making process to an artificial intelligence tool instead of independently verifying the legal authorities relied upon.
According to the Court, arbitrators have a non-delegable responsibility to personally exercise their judgment and ensure that the legal foundations of their decisions are accurate.
The decision emphasizes several core principles of arbitration, including:
- Respect for party autonomy;
- The obligation to provide properly reasoned decisions;
- Protection of the integrity of the arbitral process;
- The requirement that arbitrators personally discharge their adjudicative responsibilities.
The Court concluded that these obligations cannot be transferred to technological tools, including AI systems.
Confidence in the Arbitration Process
The Superior Court also stressed that reliance on fabricated legal authorities undermines confidence not only in the individual award but also in arbitration as a whole.
The integrity of the dispute resolution process depends upon parties having confidence that decisions are based on genuine legal analysis rather than unverified AI-generated content.
AI Is Not Prohibited
Importantly, the Court stopped short of prohibiting the use of artificial intelligence altogether.
Instead, it adopted a balanced approach, recognizing that AI can serve as a useful drafting or research aid while cautioning against unquestioning reliance on its output.
The Court acknowledged several potential risks associated with AI, including:
- Fabricated or inaccurate legal citations;
- Lack of independent legal judgment;
- Privacy and confidentiality concerns.
At the same time, it clarified that the mere use of AI does not automatically invalidate an arbitral award.
Whether an award should be set aside will depend on the particular circumstances, including:
- The seriousness of any procedural irregularity;
- The extent to which AI-generated errors influenced the decision;
- The actual impact on the fairness and validity of the award.
Minor inaccuracies or limited use of AI as a drafting tool would not necessarily justify overturning an award.
Looking Ahead
As artificial intelligence becomes increasingly integrated into legal practice and arbitration, this decision provides important guidance for arbitrators and practitioners alike.
Several arbitral institutions have already begun developing guidelines governing the responsible use of AI, but many practical and ethical questions remain unresolved.
The Superior Court’s decision demonstrates that while technology can assist legal professionals, it cannot replace the independent judgment that lies at the heart of the arbitral function.
Final Thoughts
The Quebec Superior Court’s decision serves as an important reminder that artificial intelligence is a tool—not a substitute for legal decision-making.
Although AI will likely continue to play an expanding role in arbitration and other forms of dispute resolution, the responsibility for evaluating the law, verifying authorities, and reaching a reasoned decision remains with the arbitrator.
Ultimately, the judgment reinforces a fundamental principle of arbitration: technological innovation may assist the process, but the duty to decide the dispute—and to do so on an accurate legal foundation—cannot be delegated.