The “Costs” Of Cutting Corners With AI: Mazaheri v Law Society Of Ontario

As artificial intelligence (“AI”) tools become more sophisticated and accessible, their use continues to expand across virtually every aspect of modern life. AI-powered tools now have the ability to perform tasks that once required significant human input, such as drafting documents, analyzing information, and generating content. While these technologies can offer significant benefits by enhancing productivity and increasing access to information, it is essential to consider the dangers associated with their use, particularly in the legal field.

AI systems are simply not perfect and can produce inaccurate or entirely fabricated information, including fictitious case law and misleading legal authorities. These outputs, often referred to as “hallucinations”, highlight the inherent limitations of relying on AI, particularly for legal research. Accordingly, negligent and reckless reliance on AI remains a growing concern within the legal profession because courts and other adjudicative bodies, such as tribunals, can be misled and the efficiency and integrity of the legal process can be undermined.

In Mazaheri v Law Society of Ontario, 2026 ONLSTH 112, the Law Society Tribunal Hearing Division (the “Tribunal”) ordered a lawyer (SM) to pay $31,150 in costs after finding that submissions made in support of two motions before the Tribunal contained fictitious cases and legal authorities that did not support the propositions for which they were cited. In preparing the submissions and materials for the motions, SM had used AI, in particular Grok, and had failed to verify its output.

What Happened in Mazaheri?

SM had been linked to mortgage fraud scheme involving the late Arash Missaghi. On June 17, 2024, Mr. Missaghi was killed by a victim of the fraudulent scheme. The victim then took his own life.

Subsequently, the Law Society of Ontario opened an investigation into SM and in November 2024 the Tribunal suspended his license to practice law on an interlocutory basis.

Almost a year later, in October 2025, SM brought a motion to cancel or vary the interlocutory suspension of his license and later brought a second motion to exclude evidence that the Law Society filed in response to the first motion. SM also sought to have the panel recuse itself on the grounds of alleged bias because they had seen the Law Society’s evidence.

In support of his motions, SM filed materials which contained incoherent legal arguments, numerous fictitious cases, and cases that existed but did not support the legal propositions for which they were cited.

For example, SM cited and hyperlinked the case of “Law Society of Ontario v. Mercer” which actually hyperlinked to the case of Law Society of Ontario v. Deokaran. In his factum, SM also cited paragraph 29 of Law Society of Ontario v. Marusic, 2015 ONLSTH 81 for the proposition that sanctions were imposed for procedural excesses. However, paragraph 29 of Marusic said something else.

As a result of SM’s filings, the Tribunal was required to issue a case management direction and to reschedule the hearing dates for the motions.

Ultimately, the Tribunal rejected both of SM’s motions and considered his misuse of AI in its determination of costs. While the Law Society argued that its full costs of $31,150 should be awarded, SM asked that the costs request be dismissed in its entirety, or alternatively, that costs be limited to $5,000 or less, or that the issue be deferred to the panel hearing the merits of the Law Society’s conduct application. SM offered no justification for the Tribunal to defer its costs decision to the hearing panel, and the Tribunal decided that it was in a better position to assess the appropriate costs award.

While the Tribunal noted that it would have ordered SM to pay most of the Law Society’s costs even if he had not used AI in the way that he did, SM’s irresponsible and negligent use of AI warranted an award of full costs in the amount of $31,150.

Impact on Proceeding and Cost Analysis

The Tribunal explained that SM’s materials significantly complicated what would otherwise have been relatively straightforward motions, and caused additional hearings to be scheduled to address the hallucinated authorities and misleading legal arguments. The Tribunal and opposing counsel were forced to spend considerable time identifying, documenting, and responding to the inaccuracies.

In the Tribunal's view, these unnecessary steps resulted in wasted time and increased costs.

While the Tribunal stated that the problem was not SM’s use of AI per se, the Tribunal was concerned about SM’s failure to independently verify the information generated by the technology before relying on it. The Tribunal explained that AI can fabricate content, cannot think like a lawyer, and cannot exercise the same level of legal judgment, assess the relevance of authorities, or verify the accuracy of its own output like a real lawyer.

In seeking to avoid a full costs award, SM provided a written apology to the Tribunal claiming that his “over-reliance on generative artificial intelligence” came about from trying to manage the filing of materials on his own, without a lawyer. However, this excuse was unpersuasive. The Tribunal affirmed that had he retained a lawyer, SM would still have been responsible for the materials he submitted.

Three Key Takeaways for Lawyers and Self-Represented Litigants

This decision serves as another important cautionary reminder to both lawyers and self-represented litigants to exercise independent judgment and carefully verify all AI-generated information before relying upon it. A citation should never be relied upon without confirming that the case exists and supports the proposition for which it is cited.

Furthermore, the case highlights that generative AI tools are not a replacement for legal judgment because AI tools cannot assess legal nuance, determine relevance, or exercise professional judgment like a human.

Lastly, the case demonstrates that courts and tribunals are carefully scrutinizing submissions to assess whether they contain AI-generated hallucinations, that they will not easily be fooled, and that they will tolerate its misuse.

This case represents one of the few instances of the misuse of AI by a lawyer before the lawyer’s professional regulatory body. While SM is already facing discipline for other conduct, it remains to be seen if the Law Society of Ontario will now open a separate investigation into SM’s misuse of AI in the disciplinary proceedings.

[This blog was not produced with the assistance of AI]. A PDF version is available to download here.

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.