In 1182573 B.C. Ltd. V. Gerber, 2026 BCCA 399, the Plaintiff / Appellant was self-represented and he acknowledged his reliance on AI in preparing his submissions. Noting that Courts are increasingly experiencing self-represented litigants who rely on generative AI, the Court had several comments. And though directed to self-represented parties, the issues this Court recognized are also appearing in the work product of lawyers:
[79] Our courts are increasingly experiencing self-represented litigants who have relied extensively on generative artificial intelligence (“AI”)…
[80] I have a few words of advice and warnings to give litigants about the use of AI-generated materials in court. Using generative AI tools without sufficient human oversight and caution creates a number of risks.
Risk of Hallucination
[81] The risk that AI will generate hallucinations, by making up case authorities that do not in fact exist, making up purported case quotes that do not exist, or incorrectly summarizing real cases, is by now fairly well known.
[82] A litigant who files submissions containing false case authorities will be in violation of the Registrar’s Filing Directive (4 September 2025). This Directive acknowledges the rapid development of AI and reiterates that litigants are “responsible for the authenticity and accuracy of all materials filed with the Court”: at p 14. A litigant who files submissions without reviewing them to ensure their authenticity and accuracy is in violation of this Directive and risks a costs award being issued against them: Wu v Murray, 2025 BCCA 365 at paras 11-15.
[83] If a litigant uses AI to assist them in searching for relevant legal authorities, in my view they should focus on the free legal source and search tools provided by CanLII and found at www.canlii.org. The CanLII tool uses AI to generate search terms based on plain language questions that a litigant enters. It then identifies real cases that may be relevant to the search query.
[84] No matter what tools the litigant uses, it remains their responsibility to ensure that the authorities they put before the court are real (not hallucinated) and that they are relevant to the legal or factual issues being decided.
Risk of Attaching Numerous and Irrelevant Legal Authorities
[85] Another risk of using AI is that the non-legally trained person will not exercise their own judgment as to what is or is not relevant.
[86] From my observations, it appears to me that some non-legally trained litigants are using AI to generate long lists of case authorities and then excerpting sentences or paragraphs from those authorities into their submissions as though these cases and excerpts support their arguments, without realizing (or perhaps caring) that the many cases they are citing are irrelevant to the issues before the court.
[87] It is the opposite of persuasive advocacy to cite and excerpt long lists of irrelevant case authorities. It makes submissions incoherent.
Risk of Bad Legal Advice
[88] There is an additional risk that is involved with the use of generative AI as a tool to assist in a court action. That risk is that AI will give bad legal advice and that the non-legally trained litigant will not recognize this.
[89] Again, from personal observation, it appears that increasingly the courts are receiving pleadings and submissions that are unduly lengthy and verbose, containing “legal” language untethered from established legal principles, and which are incomprehensible. These types of materials do nothing to advance a party’s legal case. It seems likely that at least some of these materials are AI generated. It appears that AI can be quite delusional about one’s legal rights and how to present legal argument.
[90] I am also concerned by the increasingly aggressive tone I am seeing in some litigation files, not just in the pleadings and submissions but also in correspondence. It causes me to wonder if AI may be leading individuals to lose their human moral instincts to be polite, measured, honest, and ethical. AI has no moral code: no sense of ethics or honesty. Poor conduct by a party within litigation not only exacerbates the conflict and leads to more expense, but it can also undermine the person’s credibility. This can contribute to the court making negative findings against them, and a high costs award against them at the end of the case.
[91] A party using AI to assist them needs to be aware of these risks, needs to carefully review and circumscribe the use of such material, and must exercise good judgment in culling it and in toning it down.
[92] Importantly, generative AI does not understand the law, legal system, legal reasoning, or exercise good judgment. Our society is complex and so too is the legal system. Understanding the law and having the judgment to know how to advance a case is complicated. That is why lawyers require years of training and experience and are professionally regulated. AI does not replace the expertise and value of trained lawyers.
[93] That is not to say I do not recognize that lawyers may also be using AI, and these same risks apply to them. But lawyers who use AI can apply their professional training and judgment to the results to guard against these same risks.
Access to Justice Concerns
[94] The use of AI to prepare court pleadings and submissions, without exercising independent and informed judgment, can result in multiple lawsuits and applications that have no merit. It can unfairly increase the time and expense of the respondents who have to respond to the materials.
[95] It can also increase the judicial resources expended on reviewing the materials. This raises access to justice concerns.
[96] Judges are a limited human resource. They must organize their time to best serve the broader public in deciding disputes based on established legal principles and evidence. Judges cannot provide an unlimited audience to litigants who have relied on AI to generate submissions that are not relevant. Judges will not always have time to explain at length that an AI-generated argument is baseless. Judges have to focus their attention on the issues that are truly legally relevant in a particular case.
[97] Simply put, a litigant who decides to use AI to help them in their case needs to be cautious, and to be aware that it might not help but rather hinder and lead them astray. They could end up disappointed and confused, as well as subject to successive costs awards.
Looking back at what we have been saying for the past three or so years on the use of, and in some cases over-reliance on, generative AI for legal matters, particularly by self-represented parties but also by lawyers, we have already said a lot of this.