As I continue to accept appointments as an arbitrator and mediator, I like to share interesting and important developments in these areas of practice which I become aware of. A recent Québec decision provides an unusually direct warning about an arbitrator’s use of artificial intelligence and addresses the integrity and enforceability of arbitral awards when these warnings are not heeded.
In Association des ressources intermédiaires d’hébergement du Québec (ARIHQ) v. Santé Québec, 2026 QCCS 1360, the Québec Superior Court annulled an award that relied upon non-existent or wrongly identified legal authorities cited by the arbitrator. Those authorities were not peripheral: they were at the heart of the arbitrator’s reasoning and were the only doctrinal or judicial authorities used as legal support for the Award. The Court concluded, on a balance of probabilities, that decision-making authority had been delegated and that the arbitrator had abdicated his responsibility to review the result.
Importantly, the Court did not hold that an arbitrator’s use of AI is itself improper. Its concern was that the parties had selected a particular arbitrator and were entitled to receive that arbitrator’s own judgment and reasons. Because the fabricated authorities lay at the heart of the reasoning, the procedural failure was sufficiently serious to undermine confidence in both the award and the arbitral process. The Court concluded that the breach probably had a significant effect on the result.
The judgment is in French; the original remains authoritative. An English case comment is linked at the end of this note.
The decision appears to be the first publicly reported judgment to annul an award on this basis. An earlier U.S. challenge, LaPaglia v. Valve Corporation, alleged that an AAA arbitrator had “outsourced his adjudicative role” to AI. That petition was dismissed for lack of federal jurisdiction, however, without deciding the AI issue.
International guidance is developing rapidly. Ciarb’s Guideline on the Use of AI in Arbitration and the SVAMC Guidelines on the Use of Artificial Intelligence in Arbitration both recognize that AI may assist an arbitrator but must not displace independent judgment. Sources, facts and analysis must be verified, confidentiality protected, and adjudicative responsibility retained by the tribunal.
The ICC has established a Task Force on Artificial Intelligence in Dispute Resolution, although it has not yet issued comparable guidance. Notably, article 44 of the 2026 ICC Arbitration Rules expressly prohibits delegating decision-making authority to a tribunal secretary, a principle equally pertinent to AI.
The practical lesson is simple: AI may be a useful assistant, but it cannot be the arbitrator.
With appropriate appreciation of the irony, generative AI assisted in researching and drafting this note under my direction. I directed and reviewed that process and checked the authorities and legal propositions against the original sources. The conclusions, and any errors, remain my own.
For further reading, an English case comment is available from Arbitration Matters.