Background
Myriane Béland, a self-represented public servant at Health Canada, sought judicial review of a July 2, 2025 decision by a Commissioner at the Public Sector Integrity Commissioner’s office (PSIC). The Commissioner had declined to deal with Béland’s reprisal complaint under s. 19.1(1) of the Public Servants Disclosure Protection Act (PSDPA), finding her complaint inadmissible and outside the Commissioner’s jurisdiction. Béland alleged patterns of exclusion, undermining, and role erosion in her workplace that caused psychological harm and professional disadvantage, characterizing these as reprisals for a protected disclosure of wrongdoing.
A preliminary issue arose two business days before the hearing, when the respondent informed the Court that Béland’s factum relied on several non-existent or inapplicable cases — apparently produced by AI “hallucinations.” Three of the five cases cited in her factum bore neutral citations that actually corresponded to different, unrelated decisions. Five further cases in her case compendium also appeared to be fictitious. Béland had never disclosed her use of AI as required by the Federal Court’s May 7, 2024 Notice on the Use of Artificial Intelligence in Court Proceedings. At the hearing, she admitted using AI to prepare her materials and stated she had been unaware of the Notice and that it was her first appearance before a court.
The Court’s Holding
Justice Azmudeh declined to strike Béland’s entire record, as the hallucinated citations affected only a few paragraphs and there was no apparent intent to mislead the Court. However, the Court found that citing non-existent or fabricated authorities — even by a self-represented litigant — is a serious matter that can mislead the court and undermine the administration of justice. Citing a growing body of Federal Court authority on AI use (including cases where costs were imposed on counsel and records were struck for similar conduct), the Court ordered Béland to pay $500 in costs for her undisclosed use of AI in violation of the AI Notice. The hallucinated citations were disregarded, with the Court noting that all AI-generated case references related specifically to Béland’s procedural fairness arguments, leaving those arguments without any supporting legal authority.
On the merits, the Court dismissed the judicial review on both grounds raised. As to reasonableness, the Commissioner’s decision was upheld: under the PSDPA, a reprisal complaint requires proof that the complainant made a protected disclosure of a wrongdoing and then suffered an adverse employment measure because of that disclosure. Béland’s only putative disclosure was a letter sent to an Assistant Deputy Minister on the same day she filed her PSIC complaint — a letter the ADM may never have seen, and which was followed by no evidence of any retaliatory measure. The Commissioner reasonably declined to give that letter any weight as a qualifying protected disclosure. As to procedural fairness, Béland had a full opportunity to present her case and supporting documents. She was in France without notifying the PSIC when the decision was rendered, and she did not identify any additional evidence she would have submitted. No promise of a particular procedure had been made, and the PSIC’s publicly available policy described the admissibility criteria she was required to meet.
Key Takeaways
- A reprisal complaint under the PSDPA requires both a prior protected disclosure of a wrongdoing and a subsequent adverse employment measure causally linked to that disclosure; a complaint that is purely personal in nature falls outside the Commissioner’s jurisdiction.
- Self-represented litigants are held to the same obligation as counsel to disclose AI-assisted preparation of court documents under the Federal Court’s AI Notice, and to verify the accuracy of any AI-generated content before filing; ignorance of the Notice is not an excuse.
- Citing hallucinated (AI-fabricated) legal authorities without disclosure is a serious breach that can result in costs sanctions, striking of pleadings, or personal costs against counsel, depending on the circumstances; the Court surveyed a growing line of Federal Court decisions imposing a range of sanctions for this conduct.
- A complainant who fails to notify a decision-maker of her absence and unavailability during the decision-making process cannot later claim a breach of procedural fairness on that basis.
Why It Matters
This decision adds to a rapidly developing body of Federal Court jurisprudence on the judicial treatment of AI-generated hallucinations in court filings. It confirms that the obligation to disclose AI use and verify citations applies equally to self-represented litigants — not just lawyers — and that the range of sanctions is calibrated to the severity of the misconduct and the degree of candour shown. The $500 costs award reflects a measured response where the litigant ultimately admitted AI use, did not appear to intend deception, and the remainder of the record was salvageable.
On the substantive side, the case reinforces the narrow jurisdictional gateway of the PSDPA’s reprisal regime. Employees experiencing workplace harassment or discrimination cannot use the PSDPA as a substitute for human rights or labour grievance mechanisms unless they can clearly establish an anterior protected disclosure of a statutory wrongdoing and a causal link to adverse treatment. The decision, building on Dempsey v. Canada (Attorney General), 2025 FC 245, underscores that complainants bear the burden of providing a clear and convincing account of those foundational elements from the outset.