Alberta courts are increasingly penalizing the misuse of artificial intelligence (AI) in legal documents. The Alberta Court of King’s Bench (ABKB) and the Alberta Court of Appeal (ABCA) have released several recent decisions where additional costs were awarded against litigants, including one award against counsel personally, for the improper use of AI. While there are no express prohibitions on the use of AI, the onus remains on litigants and counsel to ensure AI-generated legal sources are accurate. Those who fail to check the accuracy of AI-generated content in their legal documents do so at their own peril.
The 2023 Notice to the Legal Profession
In 2023, the ABCA, ABKB and the Alberta Court of Justice issued a joint Notice to the Legal Profession (the Notice) cautioning all legal practitioners, including self-represented litigants, about the use of large language models in their court submissions. The Notice urged parties to rely exclusively on authoritative sources when referencing case law, statutes or commentary, and ensure a human remains “in the loop” to verify AI-generated submissions.[1]
Alberta courts weigh in
More recently, the ABCA and ABKB have not shied away from relying on the Notice to implement additional cost awards for the misuse of AI in court documents.
Reddy v. Saroya, 2025 ABCA 322/2026 ABCA 20
In Reddy v. Saroya (Reddy), a factum was filed by the appellant’s counsel which was drafted by a third party and referenced several cases that did not exist. After being alerted to issues with locating those cases, counsel initially brushed off the concerns as merely errors in the citations.[2] When the issues were eventually discovered, the appellant was granted permission to file an amended factum, along with the respondent.[3]
The ABCA held that while ultimately left to the discretion of the individual judge or panel, maintaining the integrity and credibility of the court processes justifies imposing penalties and sanctions for the misuse of AI, which could include striking submissions or costs awards against the party or counsel.[4] Further, counsel and self-represented litigants were cautioned that they should not expect leniency when they have failed to adhere to the clear and unambiguous requirements set out in the Notice.[5]
The ABCA subsequently requested additional written submissions on whether costs against the appellant’s lead counsel were appropriate given the improper use of AI. Ultimately, the appellant’s lead counsel was ordered to pay a costs award in the amount of CA$17,550 as the ABCA found “sufficiently serious misconduct” existed that seriously interfered with the administration of justice. In doing so, the ABCA reemphasized that the lawyer whose name appears on the filed document bears responsibility for its contents.[6]
DJ v. SN, 2025 ABCA 383
In DJ v. SN, a self-represented litigant admitted to using AI in her factum, which referenced three non-existent cases.[7] Relying on the ABCA’s warning in Reddy and noting that the Notice applies equally to counsel and self-represented litigants, the ABCA awarded CA$500 in additional costs against the litigant for failing to check the accuracy of AI-generated sources in her submissions.[8] Further, the ABCA warned that “more substantial penalties” will follow in future cases of non-compliance with the Notice.[9]
Barrette v. Wevers, 2025 ABKB 723
In Barrette v. Wevers, the ABKB issued an additional costs award in the amount of CA$1,000 per set of respondents, specifically targeting a self-represented litigant’s improper use of AI in his factum, which referenced at least three fake authorities (two fake cases and one fake regulation).[10] After noting the litigant sent three sets of counsel and the ABKB on a “wild goose chase,” the ABKB found itself bound by the ABCA’s comment in DJ v. SN that “more substantial penalties” would follow in future cases of AI misuse. As a result, the ABKB was “obligated to up the ante” by awarding an increased costs award compared to DJ v. SN.[11]
Practical takeaways for using AI in court submissions
As the use of AI becomes more prevalent, Alberta courts are taking a firm stance against its misuse in court documents. In these circumstances, litigants and counsel should not expect leniency from courts, particularly in light of the ABCA’s comment that sanctions will continue to increase in severity. The best way for a litigant or legal counsel to protect themselves against violating the Notice is to verify the accuracy of AI-generated content before using it in their submissions to the Court. Those who fail to adhere to this practice may incur increased costs or other penalties at the discretion of the presiding judge or panel. As it relates to legal counsel, these recent decisions provide a healthy reminder that once your name is on a legal document, you are responsible for its contents, whether good or bad.
For more information about the proper use of AI in legal documents or the decisions and Notice discussed in this blog, please reach out to Sean Fairhurst and Brenden Roberts.[12]
[1] “Notice to the Public and Legal Profession Enduring the Integrity of Court Submissions When Using Large Language Models” (6 October 2023), online: <https://albertacourts.ca/docs/default-source/ca/tri-court-notice-to-profession-and-public—large-language-models.pdf?sfvrsn=1d1e5a82_3>
[2] Reddy v. Saroya, 2026 ABCA 20 at para 4 [Reddy 2]
[3] Ibid at para 5.
[4] Reddy v. Saroya, 2025 ABCA 322 at para 84 [Reddy 1].
[5] Ibid.
[6] Reddy 2, supra note 2 at paras 14, 16.
[7] DJ v. SN, 2025 ABCA 383 at para 24.
[8] Ibid at para 26.
[9] Ibid.
[10] Barrette v. Wevers, 2025 ABKB 723 at paras 26—27.
[11] Ibid at para 27.
[12] A special thanks to Nick Welch, student-at-law, who assisted with this blog.
