Law360 Canada (June 5, 2026, 11:02 AM EDT) --

Tom Macintosh Zheng

When a Canadian court or tribunal finds that a party has relied on a case that does not exist, the consequence is far from uniform. In one decision, the lawyer responsible was ordered to pay $17,550 in costs personally. In another, the order was $100. In 60 of the 177 decisions we reviewed, the adjudicator identified the problem but imposed no consequence at all.

In March, we reported in Law360 Canada that fictitious case law had become a systemic problem in Canadian courts and tribunals. This article turns to the response. Once an adjudicator identifies that a party has cited cases that do not exist, what follows? To answer that, we tracked every publicly available Canadian decision between February 2024 and June 2026 that flagged a fictitious citation or another form of AI misuse and recorded the outcome in each. The dataset now stands at 177 decisions across 49 courts and tribunals, and the pattern it reveals is one of marked inconsistency.

What we found

Of the 177 decisions, 151 (85.3 per cent) involved fictitious citations, that is, cases cited as authority that do not exist. The remaining 26 (14.7 per cent) involved other forms of AI misuse, including AI-hallucinated propositions of law and fabricated quotations from real cases.

The responses fell into three categories. In 60 decisions (33.9 per cent), the court or tribunal identified the problem but imposed no sanction. In 82 (46.3 per cent), it issued a warning. In 35 (19.8 per cent), it imposed a serious sanction: a costs order, a regulatory referral, contempt proceedings, the striking of a filing, or a comparable consequence.

The gap between courts and tribunals

The clearest pattern in the data is that courts punish fake law far more readily than tribunals do.

Courts presided over by judges or associate judges, including the superior courts, courts of appeal, the Federal Court and the provincial courts, accounted for 74 of the decisions and imposed a serious sanction in 37.8 per cent of them. Administrative tribunals and boards accounted for the remaining 103 decisions but imposed a serious sanction in only 6.8 per cent.

A court was, therefore, more than five times as likely as a tribunal to impose a serious consequence for the same conduct. The divide is just as stark at the other end: tribunals left the conduct unsanctioned in 46.6 per cent of cases, against 16.2 per cent in the courts. Warnings were common in both, at 45.9 per cent in courts and 46.6 per cent at tribunals.

The role of representation

Whether a party was represented also shaped the outcome. Self-represented litigants accounted for 84.7 per cent of the decisions but drew a serious sanction in 15.3 per cent of their cases. Represented parties accounted for 15.3 per cent of the decisions but drew a serious sanction in 44.4 per cent of theirs. Where a lawyer was responsible for the fictitious material, the consequence was more often a serious one, and frequently a personal one: five of the 16 costs orders, totalling $18,450, were paid by counsel rather than the client.

Why the response varies

Some of this variation reflects deliberate judgment rather than oversight. Administrative tribunals carry heavy caseloads. An adjudicator confronted with submissions that mix sound arguments with unverifiable ones may reasonably decide the matter on the arguments that hold and disregard the rest, without pausing to address the citations that do not exist.

There is also a practical evidentiary hurdle. Concluding that a fictitious citation was generated by an AI tool requires a factual finding. Doing so involves a more resource-intensive exercise, one that a busy tribunal may have little reason to undertake when the underlying matter can be resolved without it.

The case for consistency

Even allowing for these reasons, the cumulative picture is difficult to reconcile. Two litigants can engage in materially the same conduct, each relying on a case that does not exist, and one will face no consequences while the other is named in a published decision, warned on the record or ordered to pay costs. When the difference in treatment turns less on what the litigant did than on the forum they happened to be in, the inconsistency begins to raise a question of fairness.

The answer is not necessarily harsher sanctions. A self-represented litigant who unknowingly relied on a flawed tool is not in the same position as experienced counsel, and the response should reflect that difference. The answer is consistency: a shared understanding, across courts and tribunals, of when fabricated authority should be identified, when it should be addressed on the record, and what should follow from that point. Litigants, and the lawyers who advise them, should be able to anticipate how a Canadian adjudicator will treat fake law. At present, they cannot.

Tom Macintosh Zheng is a Toronto-based former commercial litigator. He is now building online tools to help Canadians access and understand our justice system. To see the full database tracking AI-hallucinated cases in Canadian courts, which is updated weekly, click here.

The opinions expressed are those of the author(s) and do not necessarily reflect the views of the author’s firm, its clients, LexisNexis Canada, Law360 Canada or any of its or their respective affiliates. This article is for general information purposes and is not intended to be and should not be taken as legal advice.

Interested in writing for us? To learn more about how you can add your voice to Law360 Canada, contact Analysis Editor Richard Skinulis at Richard.Skinulis@lexisnexis.ca or call 437-828-6772.