Adapting to AI: Continuing Developments in How Canadian Courts Are Responding to Generative Technologies
Canadian Courts, administrative tribunal bodies, and law societies continue to move to respond to the rise in the use of generative artificial intelligence tools, like ChatGPT, in litigation. This push has been driven, in part, by an increase in the number of submitted briefs containing fictitious or ‘hallucinated’ AI generated legal citations.
This article was originally published in the Toronto Law Journal.
Declaration Regimes Adopted
As 2023 was coming to a close, the Federal Court issued a Notice to the Parties and the Profession on the use of artificial intelligence in court proceedings.[1] Consistent with several other provincial courts, the Notice required parties to inform the Court, and the other parties, if they have used artificial intelligence to create or generate new content in preparing a document that is filed with the Court. If any such content had been included in a filed document, the Notice provided that the first paragraph of the text in that document must disclose that AI had been used.
The Notice also advised that counsel and the parties are to verify any AI-related content – that there was a “human in the loop” – and set out principles to guide the use of AI in court documents, including the use of only well-recognised and reliable sources when referring to legal authority or analysis.
This Notice was further updated in May 2024.[2] It clarified that a Declaration is not necessary where the AI program used lacks the creative ability to generate new content (such as system automation, voice recognition, or document editing), or where it was used to merely suggest changes, provide recommendations, or critique content already created by a human. It also confirmed that the inclusion of a Declaration will not in and of itself attract an adverse inference and that, when taking over a matter from another lawyer or a self-represented litigant, counsel was expected to use best efforts to provide a Declaration in respect of any previously filed materials they have reason to believe may include AI generated content.
More recently, some tribunals have followed suit by issuing their own practice notices on the use of generative AI. In April 2025, Tribunals Ontario issued its Practice Direction on the Use of Artificial Intelligence (AI) in Tribunal Proceedings[3], which largely mimics that of several Canadian courts, including the Federal Court. A similar practice notice was published in early June 2025 by the Trademarks Opposition Board.[4] The latter comes after two separate instances of hallucinated citations in matters before the Board.[5]
In addition to these practice notices, the Ontario Civil Rules Committee went even a step further. Notably, new provisions in the Rules of Civil Procedure were enacted in 2024 requiring that any factum or expert report filed with the Court include a signed statement certifying the authenticity of every authority cited therein.[6]
Hallucinations Have Occurred
Despite these measures, there have been cases in the past few years involving “hallucinated” legal citations being relied upon by lawyers.[7] The decisions in these cases reflect a commitment to balancing firmness with fairness, while making clear that there needs to be consequences for the misconduct.
In February 2024, in Zhang v. Chen[8], the British Columbia Supreme Court held counsel personally liable for costs incurred by the other party, due to the citation of two fictitious cases that had been imagined by ChatGPT. The successful party asked for special costs, which requires “a serious abuse of the judicial system… or dishonest or malicious misconduct… that is deliberate”.[9] A mere “mistake, error in judgment or even negligence” does not, however, warrant such an order.[10] While the Court does refer to the lawyer’s conduct as being an abuse of process, it ultimately found that the lawyer did not have the intention to deceive or misdirect.
The Court concluded that the lawyer should be held personally liable to pay part of the costs owed by their client to the successful party; this remedy is available where a lawyer “has caused costs to be incurred without reasonable cause, or has caused costs to be wasted through delay, neglect or some other fault”.[11] Costs were allowed for a total of 2 hearing days, as well as any reasonable disbursements. The lawyer was also ordered to review all their other matters before the court to verify whether any materials contained case citations or summaries that were obtained from ChatGPT or other generative AI tools, and to advise the Court and opposing parties accordingly.
In May 2025, a similar issue arose in Ko v. Li[12], a matter heard by the Ontario Superior Court of Justice, where Applicant’s counsel had submitted a factum citing several decisions that were either non-existent or irrelevant to the arguments and principles they purported to support. Upon realising that the case law was most likely an AI hallucination, the Court ordered applicant’s counsel to a “show cause” hearing. The Court emphasised the seriousness of these actions, noting that it is a litigation lawyer’s “most fundamental duty” not to mislead the Court.[13] Furthermore, the decision in Zhang v. Chen was distinguished as, unlike in Zhang, counsel in this case did not catch their mistake before the hearing, had not withdrawn their factum before the hearing, had not apologised and had not corrected the mistake prior to the issuance of the Court’s judgment.
Fortunately for counsel in Ko v. Li, the show cause hearing never took place. At the show cause scheduling conference, the Court withdrew the show cause order and deemed it satisfied. [14] The Court did so, in part, because counsel had admitted the facts, apologised and proposed additional positive steps, including an undertaking to complete six hours of Continuing Professional Development in legal ethics and technology. The Court also took notice that counsel had gained a certain notoriety because of their misstep, including through the publication of various articles in both the legal and general press, and accordingly concluded that the objective of the show cause hearing, including maintaining the dignity and fairness of the civil justice system and denouncing this type of misconduct, had already been met.
As judicial guidance around generative AI continues to evolve, Courts appear to be exhibiting less tolerance in cases marked by misuse. In June 2025, the Federal Court raised serious concerns in Hussein v. Canada (Immigration, Refugees and Citizenship) over the undeclared use of generative AI in court submissions, but took a more direct approach to the consequences for the misuse it faced.[15] The Applicants’ counsel failed to disclose their use of AI and submitted fabricated citations, including a hallucinated legal test, with no human verification.[16] Disclosure only came after repeated Court directions and as such, the Court found the failure to declare AI use amounted to an attempt to mislead.[17]
The Court ultimately found that a modest personal costs award should be imposed against counsel, despite the unusual nature and high threshold required for this type of sanction. While the Court found that much of counsel’s submissions mitigated the situation, the personal costs award was seen as appropriate due to the unnecessary costs and procedural complications caused by the repeated failure to disclose AI use and reliance on hallucinated authorities.[18] Interestingly, the Court affirmed that, “the real issue is not the use of generative artificial intelligence but the failure to declare that use.”[19] This contrasts with the decision in Ko v. Li, given that, unlike the Federal Court’s Notice to the Parties and the Profession, the Ontario Rules of Civil Procedure do not explicitly require disclosure of generative AI use, but rather certification of the authenticity of authorities cited.[20]
In July 2025, the Federal Court took a firm stance in Lloyd’s Register Canada Ltd. v. Choi against Munchang Choi, a self-represented litigant who submitted a Motion Record containing legal citations generated by AI and without the required declaration.[21] Choi had a documented history of relying on AI-generated false authorities, having previously misrepresented over 30 decisions in a proceeding before the Canada Industrial Relations Board.[22] The Court found his explanations of inexperience and mental health challenges insufficient, given that all parties, lawyers and self-represented litigants alike, bear full responsibility for verifying the accuracy of their filings and compliance with the AI-related disclosure requirements. The Court ordered the Motion Record to be removed from the file to preserve the integrity of the proceeding and awarded modest costs to the Applicant, deeming Choi’s conduct to be “scandalous, frivolous, vexatious, and otherwise an abuse of process.” [23]
In August 2025, the Ontario Superior Court of Justice faced a hallucination that appeared to be “too precisely on point”,[24] and pondered whether the legal submissions “were not purely the product of AI, but may have included some express manipulation”.[25] Concerned that this was not the first instance of the defendant relying on a hallucination, and, as well, that if the Court was to dismiss the defendant’s motion, a new motion would likely follow, the Court considered and then advised that a prohibition on future motions would come without further warning if a hallucination was submitted again.[26]
In July 2025, the Federal Court took a firm stance in Lloyd’s Register Canada Ltd. v. Choi against Munchang Choi, a self-represented litigant who submitted a Motion Record containing legal citations generated by AI and without the required declaration.[21] Choi had a documented history of relying on AI-generated false authorities, having previously misrepresented over 30 decisions in a proceeding before the Canada Industrial Relations Board.[22] The Court found his explanations of inexperience and mental health challenges insufficient, given that all parties, lawyers and self-represented litigants alike, bear full responsibility for verifying the accuracy of their filings and compliance with the AI-related disclosure requirements. The Court ordered the Motion Record to be removed from the file to preserve the integrity of the proceeding and awarded modest costs to the Applicant, deeming Choi’s conduct to be “scandalous, frivolous, vexatious, and otherwise an abuse of process.” [23]
It’s Not Just Canada
A Thomson Reuters study, drawing from its Westlaw database, showed a concerning number of hallucinations and citations of non-existent legal cases in U.S. Courts. The study found 22 different cases in one month – July 2025 – where the Courts or opposing parties had identified non-existent cases within Court filings.[27]
In a decision published in July 2025, the Federal Court of Australia ordered Massar Briggs Law to pay costs after a junior solicitor submitted court materials containing AI-fabricated citations sourced from Google Scholar.[28] Justice Murphy highlighted that generative AI’s tendency to hallucinate or fabricate information is now widely recognised, and accordingly, lawyers are expected to exercise vigilance and fully acknowledge the limitations of these tools.[29]
Troubling, But Infrequent
The appearance of fabricated case citations in Canadian Courts and tribunals is troubling, but as our Courts and tribunals have developed rules and guidance around the use of generative AI, as legal professionals have gained a deeper understanding of generative AI’s limitations, and as the risks and consequences of misuse have become clearer, such occurrences, while likely to continue, should, at least for legal professionals, remain infrequent.
Cara Parisien and Scott MacKendrick are qualified lawyers and IP litigation experts based in Montreal and Toronto, Canada. Access their contact information on the ROBIC website.
[1] Federal Court, Notice to the Parties and the Profession, “The Use of Artificial Intelligence in Court Proceedings”, December 20, 2023.
[2] Federal Court, Notice to the Parties and the Profession, “The Use of Artificial Intelligence in Court Proceedings”, May 7, 2024.
[3] Tribunals Ontario, “Practice Direction on the Use of Artificial Intelligence (AI) in Tribunal Proceedings”, April 2025.
[4] Trademarks Opposition Board, “Use of AI in proceedings before the Trademarks Opposition Board”, June 4, 2025.
[5] See Industria de Diseño Textil, S.A. v Sara Ghassai, 2024 TMOB 150 and Monster Energy Company v Pacific Smoke International Inc., 2024 TMOB 211.
[6] Rules of Civil Procedure, R.R.O. 1990, Reg. 194, Rules 4.06.1 (2.1) and 53.03 (6.1)
[7] Damien Charlotin, a senior research fellow at HEC Paris, has been tracking cases where generative AI hallucinated content has arisen. To-date, 27 cases in Canadian Courts and tribunals have been tracked: https://www.damiencharlotin.com/hallucinations/?q=&sort_by=-date&states=Canada&period_idx=0.
[8] 2024 BCSC 285 [Zhang].
[9] Zhang, supra. note 7, para. 26
[10] Id.
[11] Supreme Court Family Rules, B.C. Reg. 169/2009, Rule 16-1(3); Zhang, supra. note 7, par. 41.
[13] Id., para. 22.
[14] Ko v. Li, 2025 ONSC 2965.
[15] Hussein v. Canada (Immigration, Refugees and Citizenship), 2025 FC 1060
[16] Id., para 40.
[17] Id., para 41.
[18] Hussein v. Canada (Immigration, Refugees and Citizenship), 2025 FC 1138
[19] Id., para 15.
[20] Rules of Civil Procedure, RRO 1990, Reg 194, s. 4.06.1(2.1).
[21] Lloyd’s Register Canada Ltd. v. Choi, 2025 FC 1233; Federal Court, Notice to the Parties and the Profession, supra., note 2.
[22] Choi v Lloyd’s Register Canada Limited, 2024 CIRB 1146, at para. 73-79.
[23] Id.
[24] Halton (Regional Municipality) v. Rewa et al., 2025 ONSC 4503, at paras 44 & 49.
[25] Id., para. 49.
[26] Id., para 54.
[27] GenAI hallucinations are still pervasive in legal filings, but better lawyering is the cure, Thomson Reuters, August 18, 2025, https://www.thomsonreuters.com/en-us/posts/technology/genai-hallucinations/
[28] Murray on behalf of the Wamba Wemba Native Title Claim Group v State of Victoria [2025] FCA 731.
[29] Id., para. 8 & 12.
Sign up to our free newsletter to stay informed, connected and inspired by all things IP.
Featuring
