When AI Gets It Wrong: Lessons from Mazaheri v Law Society of Ontario

In the world of litigation, AI tools have become both a beacon of efficiency and source of controversy. But, the case of Mazaheri v Law Society of Ontario, 2025 ONLSTH 186 serves as a  cautionary tale and a valuable reminder that lawyers and the critical review of legal authorities are not replaceable.

In 2024, Law Society of Ontario v Mazaheri and Yack, 2024 ONLSTH 132, the LSO found that Mazaheri was wilfully blind to the fraud of his client, an oversight serious enough to warrant an interlocutory suspension of his license to practice law. The investigation continues, but Mazaheri’s suspension remains in place.

Mazaheri, undeterred, subsequently brought a motion to vary or remove the interlocutory suspension. In November 2025, at a proceedings management conference, Mazaheri objected to the admissibility of exhibits attached to an affidavit submitted by the LSO. The Tribunal directed him to file a notice of motion, affidavit, and factum in support of his objection. Rather than comply with these directions, Mazaheri filed additional documents, including a supplementary factum and affidavit. All of the materials filed were drafted with the help of generative AI.

And therein lay the problem. Not a single legal proposition was supported by a reliable authority. The documents contained a number of “hallucinations”, non-existent or misleading authorities. The parties were directed to attend a case management conference, at which a 15-page chart identifying the hallucinations contained in Mazaheri’s materials was presented.

Mazaheri admitted to using generative AI and undertook not to use such tools in the future. While the consequences of this conduct remain to be determined, Mazaheri is able to consult counsel and prepare for the next stage of the proceedings. The admissibility and bias motion were dismissed, with costs to be decided.

Although many users report positive experiences with AI, this case begs the question of whether AI is really that bad. The answer, one we may prefer not to hear, remains the same: it depends. While AI use can promote access to justice, diligent review and independent verification of AI-generated content are critical to the administration of justice. In an era of rapid technological advancements and increasingly pervasive AI use, it is important to remember that while humans are subject to error, so too is AI.