ARTICLE
7 September 2026

Fictitious Cases, Real Consequences: The Cost Of Negligent AI Use

DA
DMG Advocates

Contributor

DMG Advocates is a Toronto-based litigation boutique specializing in a broad range of disputes, including product liability, shareholder disputes, civil fraud, defamation, human rights and employment litigation and commercial arbitration. DMG’s lawyers represent both corporations and individuals, providing strategic, results-driven advocacy across a wide range of industries.
In Mazaheri v Law Society of Ontario, 2026 ONLSTH 112, the Law Society Tribunal awarded full costs, in the amount of $31,150, against the respondent taking into consideration his misuse of artificial intelligence generated motion materials.
Canada Ontario Litigation, Mediation & Arbitration
DMG Advocates are most popular:
  • within Insolvency/Bankruptcy/Re-Structuring and Employment and HR topic(s)
  • with Senior Company Executives, HR and Finance and Tax Executives
  • with readers working within the Banking & Credit, Business & Consumer Services and Healthcare industries

In Mazaheri v Law Society of Ontario, 2026 ONLSTH 112, the Law Society Tribunal awarded full costs, in the amount of $31,150, against the respondent taking into consideration his misuse of artificial intelligence generated motion materials. The decision provides a snapshot of the current state of artificial intelligence use in litigation and creates a necessary caution to legal professionals and self-represented litigants about the real consequences of AI misuse.

Factual Background

On September 23, 2025, Mazaheri brought a motion to vary or remove the interlocutory suspension of his law license.1 At the proceedings management conference on November 10, 2025, along with a notice of motion, affidavit, and factum, Mazaheri filed supplementary materials.2 Each of the materials Mazaheri produced at this conference were prepared using generative AI software that hallucinated every proposition of law.3

After these hallucinations came to the Tribunal’s attention, the Tribunal directed the parties to appear at a case management conference. This direction included a 15-page chart of examples of the hallucinated authorities.4 On November 30, 2025, Mazaheri wrote to the Tribunal admitting that he failed to verify the materials prior to submission, apologized, and undertook not to use generative AI in any further preparation.5

On December 30, 2025, the Tribunal heard Mazaheri’s motion to vary the interlocutory suspension of his law license along with admissibility and bias motions. The admissibility and bias motions were dismissed.6 The Tribunal did not address Mazaheri’s negligent use of AI since he failed to satisfy the first step of the test to cancel or vary an interlocutory suspension. However, the Tribunal noted that Mazaheri’s use of AI has impacted his credibility,7 which made it less likely that the order would be varied or cancelled.8 The Tribunal stated that they would consider Mazaheri’s use of AI when addressing costs.9

In its June 12, 2026 decision, the Tribunal ordered Mazaheri to pay $31,150 in costs to the Law Society.10 After outlining the usual factors in assessing costs, the Tribunal turned to what it called “an unusual one” –Mazaheri’s negligent use of AI in preparing materials in the admissibility and bias motions. That usage served as an additional and significantly aggravating factor in the Tribunal’s cost calculations.11

Key Findings

 AI can’t think like a lawyer

The Tribunal highlighted the rising number of reported cases on CanLII with hallucinated AI usage: 7 in 2024, 86 in 2025, and 39 in the first quarter of 2026.12 While most of these cases involved self-represented litigants, surprisingly, 24 of 132 cases involved represented parties.13

Despite its rapid growth, unlike humans, large language models of generative AI are unable to appreciate nuance, exercise judgment, or use a moral compass.14 The Tribunal noted that while a large language model can identify issues and arguments correctly, it lacks judgment and thus does not question or check its own work. Unlike a human lawyer, generative AI can’t verify whether a case it cites in support of an argument in fact supports that argument, nor can it critically assess a factual matrix to ascertain whether the law it cites apply to the facts in dispute. Importantly, the Tribunal held that unlike humans, a large language model does not know the limits of its knowledge.15 The Tribunal was clear that it was not finding that lawyers should not use AI, but rather “to state the obvious – they should use it responsibly”.16

Incorrect use of AI is costly

Despite not finding that Mazaheri intended to deceive the Tribunal, the Tribunal determined that Mazaheri’s misuse of AI was deceptive in both substance and tone due to the overconfident misstatements of law.17

The cost decision paid much consideration to the distraction and delay that came with Mazhari’s AI submissions. As a result of Mazhari’s negligence, the Tribunal had to go through the process of verifying every citation, documenting inaccuracies, and deciding how to proceed on an issue completely unrelated to the case before them.18 Such distraction resulted in the cancellation of scheduled hearings and necessitated two case management sittings to address the defective materials, all of which would have been avoided if the respondent had used AI responsibly. The Tribunal characterized the respondent’s actions as a “complete lack of respect for the administration of justice.”19

The Tribunal cited several cases where courts have awarded costs for the irresponsible use of AI.20 These cases imposed what the Tribunal called the “strong end of the spectrum”, namely, solicitor-client costs or costs against a lawyer personally.21 Cost sanctions will be used by decision makers to denounce the irresponsibly use of generative AI.

Implications

AI is changing the landscape of the legal profession. When used responsibly, it can help to drive efficiencies from which lawyers and clients benefit. But those efficiencies evaporate if lawyers (and litigants) do not apply human judgment, critical thinking skills and a moral compass to generative AI’s work product. AI is a tool, not a lawyer.

Footnotes

1. Mazaheri v Law Society of Ontario, 2025 ONLSTH 186 at para 6 [186].

2. 186 at para 8.

3. 186 at para 10.

4. 186 at para 11.

5. 186 at para 12.

6. 186 at para 45.

7. Mazaheri v Law Society of Ontario, 2026 ONLSTH 33 at para 63 [33].

8. 33 at para 67.

9. 33 at para 68.

10. Mazaheri v Law Society of Ontario, 2026 ONLSTH 112 at para 86 [112].

11. 112 at para 34.

12. 112 at para 36.

13. 112 at para 37.

14. 112 at para 44.

15. 112 at para 46.

16. 112 at para 49.

17. 112 at para 69.

18. 112 at para 74.

19. 112 at para 77.

20. 112 at para 80.

21. 112 at para 81.

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]
See More Popular Content From

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