In its decision in ARIHQ v. Santé Québec (2026 QCCS 1360), the Québec Superior Court issued one of the most significant judicial rulings to date regarding the use of generative artificial intelligence (AI) in drafting an arbitral award. The court set aside the arbitral award because the arbitrator had impermissibly used generative AI in drafting it.
The facts
The subject matter of the arbitration proceedings was a claim for payment in the millions by a clinic against a public healthcare institution. The arbitral tribunal dismissed the claimants’ case. In the subsequent motion to set aside the award filed with the competent Canadian court, the claimants demonstrated that the written reasoning in the arbitral award cited numerous purported court decisions and legal sources that did not actually exist. All legal sources on which the arbitral award was based were the result of a hallucinating AI.
The court’s decision
The competent Canadian state court set aside the arbitral award in a decision dated 22 April 2026. In the court’s view, there is no general prohibition on the use of AI systems in the context of legal work. In particular, their use is permissible to assist with research, the structuring of arguments, or the drafting of texts. However, such use becomes impermissible if the actual legal analysis or decision-making is effectively delegated to an AI system, or if AI-generated content is adopted without careful review. The court essentially relies on two principles here.
1. Parties’ autonomy to choose their own decision-maker: In arbitration proceedings, it is within the parties’ autonomy to choose their arbitrator. The parties to an arbitration proceeding select arbitrators for various reasons, including legal expertise and experience. The parties are therefore entitled to expect that the arbitrator(s) they have selected will hear and decide the dispute.
2. Personal reasoning of the award: Just as the parties are entitled to expect that the arbitrator(s) they have selected will hear and decide their case, they are also entitled to receive a personalized explanation of the decision. The court emphasizes that written reasons for decisions contribute significantly to transparency and trust in the judicial process.
Implications for the future
The court expressly noted that not every arbitral award drafted with the aid of AI would be set aside. If AI is used only minimally or in a secondary capacity, a decision may stand. However, if, as in the present case, legal sources that do not actually exist form the core of the reasoning behind the decision, the violation is serious enough to undermine confidence in the arbitral award and in the arbitration process in general.
Conclusion
In ARIHQ v. Santé Québec (2026 QCCS 1360), the Québec Superior Court sets an important precedent for the responsible use of artificial intelligence in the legal field. The ruling confirms that while modern AI applications can support legal work, the legal assessment and responsibility for decision-making remain exclusively with humans.
