Steven Hinkley (Alberta Court of Justice) has posted Courts, Regulators, and the Hallucination Problem: How Canadian Law Is Responding to Large Language Models in Legal Submissions (Alberta Law Review, forthcoming) on SSRN. Here is the abstract:
The rapid integration of Large Language Models (LLMs) into the Canadian legal profession has introduced significant challenges to the administration of justice, most notably the phenomenon of artificial intelligence “hallucinations” resulting in the submission of fabricated legal authorities to courts. This article provides a comprehensive analysis of how Canadian courts and regulators in common law jurisdictions have responded to this disruption over a critical three-year period following the release of ChatGPT. Examining the evolving regulatory landscape, the article traces the development of early judicial interventions, including the Alberta Tri-Court Notice and the Federal Court’s distinct mandatory disclosure regime, alongside guidance documents published by various provincial Law Societies and the Canadian Bar Association. It further reviews pivotal Canadian jurisprudence—such as Zhang v. Chen, Reddy v. Saroya, and Ko v. Li—to illustrate a discernible shift from initial judicial leniency toward increasingly stringent enforcement mechanisms, including personal cost awards and contempt proceedings. The analysis identifies core systemic themes, particularly the universal imposition of a non-delegable duty of human verification (“human in the loop”) anchored within existing professional obligations of competence, confidentiality, and candour. Finally, the article highlights critical under-examined tensions within the current framework, including the serious access-to-justice implications of holding self-represented litigants to strict verification standards without institutional support, and the regulatory distinction between free consumer chatbots and enterprise-grade legal AI platforms. Ultimately, the author argues that while Canada’s regulatory posture favours managed technological integration over prohibition, the enduring efficacy of relying on traditional ethical frameworks to govern rapidly advancing generative tools remains an open question for the justice system.
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.