ARTICLE
12 August 2026

Delegating To AI: Is There A Limit?

RG
Roper Greyell LLP – Employment and Labour Lawyers

Contributor

Roper Greyell LLP is one of the most well-respected and recognized workplace law practices in Canada. Our clients benefit from a team of 37 diverse and talented lawyers who are committed to providing them with the highest quality legal representation and strategic counsel in all areas of workplace law.
This page appears to be a security verification checkpoint that temporarily blocks access while validating the incoming request. The system implements protective measures to distinguish legitimate users from automated traffic or potential security threats.
Canada Media, Telecoms, IT, Entertainment
Roper Greyell LLP – Employment and Labour Lawyers are most popular:
  • with Senior Company Executives, HR and Finance and Tax Executives
  • with readers working within the Accounting & Consultancy, Basic Industries and Business & Consumer Services industries

Previously printed in the LexisNexis Labour Notes Newsletter.

In Association des ressources intermédiaires d’hébergement du Québec (L’ARIHQ) c. Santé Québec – Centre intégré universitaire de santé et de services sociaux du Centre-Sud de L’Ile-de-Montréal et 9342-4935 Québec Inc., personne morale faisant affaire sous le nom de Centre de Santé Osman et ME Michel A. Jeanniot, 2026 QCCS 1360, the Quebec Supreme Court vacated an arbitral award after the arbitrator used AI to write the decision, relying on hallucinated authorities.

Background

The underlying dispute related to a healthcare contract. The arbitrator rendered a decision in the contract dispute and relied in the legal reasoning on numerous authorities. The losing party considered the decision and, in doing so, tried to review the cited authorities but was unable to locate the same. It was determined that each authority cited in the decision was either invented by AI (hallucinated) or misattributed (the case citation led to an unrelated decision). Notably, the authorities on which the arbitrator had relied were not argued by the parties in the arbitration process.

In light of the above, the losing party applied to have the arbitrator’s award vacated as: (a) it violated public policy; and (b) the arbitral procedure had been violated because the arbitrator’s role as decision-maker had been delegated to AI.

What is AI Hallucination?

AI hallucinations occur when a large language model (LLM), which is usually generative AI, produces an output that does not exist. In this context, the AI tool used by the arbitrator generated various case citations and authorities to support the arbitral decision but the cases, or at least their application to the subject matter of the arbitration, were made up.

Decision

Justice Sheehan of the Quebec Supreme Court confirmed that a contextual analysis is required when considering whether the use of AI in an award should result in its annulment. The three considerations include:

  1. The nature of the breach;
  2. Whether the breach affected the integrity of the proceedings; and
  3. The impact of the breach on the award.

The Court concluded that the crux of the arbitrator’s reasoning in the award relied on hallucinated authorities.1 Moreover, those authorities were the only doctrinal or case references used as legal support for the award. The other case law referenced was included in sections of the award which summarized the parties’ positions – it did not form a part of the arbitrator’s legal reasoning.

The following are particularly poignant excerpts from the decision2 that articulate the Court’s position on the use of AI in judicial proceedings:

  1. Thus, if there is nothing inherently objectionable about using a reliable artificial intelligence tool, the rules in force grant lawyers and parties an obligation of oversight aimed at ensuring the accuracy of their procedures. A lawyer who fails to carry out checks shirks their responsibilities.
  1. The preponderance of evidence therefore leads to the conclusion that the Adjudicator’s authority was delegated and that he abdicated his role of reviewing the result.

In the result, the Court ordered that the arbitrator’s award be annulled or vacated.

Takeaways

It is clear that AI is becoming more prevalent and integrated into everyday life. However, the AI tools at our use and disposal to assist with productivity and efficiency are flawed and there is a continued need for human oversight. The more we start to rely on AI without human oversight, the more we can be subject to new problems like hallucinations or misattributions.

Employers should ensure that they have an AI policy in place that sets clear boundaries with respect to the use of AI in the workplace. Any such AI policy should contemplate the use of AI by both the employer and employees.

Footnotes

1. Para. 113. – Translated using Google Translate.

2. Again translated using Google Translate.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.

[View Source]

Mondaq uses cookies on this website. By using our website you agree to our use of cookies as set out in our Privacy Policy.

Learn More