Kapahi v. Elite Real Estate: Seven Fabricated Quotations in Factum — Counsel Referred to Law Society of Ontario

Background: A Real Estate Club Arbitration and a Shocking Reply Factum

The underlying dispute involved a real estate club corporation and a related arbitral cost enforcement proceeding. Justice Myers had previously dismissed the Respondents’ motion to vary an earlier order by Justice Steele. When costs submissions were being prepared, the Applicants flagged “apparent AI hallucinations in the Reply Factum delivered on the motion to vary” by Mr. Parvaiz.

Justice Myers took the highly unusual step of personally confirming the problem: on January 28, 2026, the court sent an email to all counsel advising that “Justice Myers has confirmed that the Reply Factum delivered by Mr. Parvaiz on the motion contained quotations that did not originate from the cases to which they were attributed by Mr. Parvaiz.” The court also inquired whether counsel had used generative AI to draft the factum — and why the AI certification required under Ontario Rule 4.06.1(2.1) was absent.

The AI Issue: Seven Fabricated Quotations — All From Real Cases

Justice Myers documented seven specific paragraphs in Mr. Parvaiz’s Reply Factum where quotations were attributed to real cases with correct CanLII neutral citations — but the quotations did not exist in those cases. Examples:

• Para. 29 of the factum cited Pita Royale Inc. for a quotation about fair notice and personal liability — “Nothing like this quotation appears in the case. It is wholly made up.”
• Para. 30 cited BH Frontier Solutions Inc. v. 11054660 Canada Inc. (a real Court of Appeal case) for a quotation about the threshold for piercing the corporate veil — “Nothing like this quotation appears in the case. It is wholly made up.”
• Para. 52 cited Trustees of the Millwright Regional Council of Ontario Pension Trust Fund v. Celestica Inc., attributed to “Justice Mosley” — but the decision was written by Perell J., and the bulk of the quoted text does not appear in the case.

When confronted, Mr. Parvaiz denied using AI. He attributed the false quotations to “a lack of due care,” “human errors,” “misreading the cases cited,” “carelessness,” and “inadvertence.” Justice Myers found this explanation incomprehensible: “If he did not use AI, how did he come to make up seven paragraphs and call them quotations from real cases?”

Holdings and Disposition

1. Seven fabricated quotations documented. Justice Myers personally verified each quoted passage against the cited cases and found that all seven were “wholly made up” — not variations, not paraphrases, but entirely fabricated text attributed to real cases.
2. Denial of AI use deemed implausible. The judge could not understand how seven distinct fabricated quotations could arise from human error, misreading, or inadvertence. He framed the dilemma starkly: “Either Mr. Parvaiz used AI and has been untruthful about it, or he made up seven fake paragraphs and chose to present them as actual quotations from precedent cases.”
3. “Cover-up may be worse than the initial error.” Justice Myers echoed the classic legal maxim — applicable from Watergate to bar misconduct proceedings — that denial of wrongdoing compounds the original harm.
4. Referral to Law Society of Ontario. The court referred the decision to the Law Society of Ontario for consideration of whether disciplinary action should be taken against Mr. Parvaiz. This is one of the most severe judicial consequences short of direct contempt proceedings in the AI hallucination case law.
5. Prior Rule 57.07 costs motion had already occurred. A motion for personal costs against Mr. Parvaiz under Rule 57.07 (personal liability of solicitors for costs) had already been heard, meaning the counsel had faced financial jeopardy before the Law Society referral was made.

“Either Mr. Parvaiz used AI and has been untruthful about it, or he made up seven fake paragraphs and chose to present them as actual quotations from precedent cases… As is often the case, if Mr. Parvaiz has not been truthful, the cover-up may be worse than the initial error.”

— Justice F.L. Myers, Ontario Superior Court (Commercial List), 2026 ONSC 1438

India Angle: When Denial Makes Things Worse — LSO Referral Parallel to BCI Action

The Kapahi v. Elite Real Estate case is the most serious AI hallucination matter involving an Indian-origin lawyer (Mr. Parvaiz) in the published record. It illustrates a pattern that Indian advocates must understand: denial of AI use, when the court has verified that fabricated quotations exist, converts a professional misconduct issue into a potential dishonesty issue. The Law Society of Ontario’s jurisdiction over dishonesty parallels the Bar Council of India’s jurisdiction under Advocates Act 1961 Section 35.

Relevant Indian Law

• Advocates Act 1961 Section 35 (Professional Misconduct): The Bar Council of India State Bar Councils have jurisdiction over “professional misconduct” by advocates. Fabricating legal quotations and submitting them as part of a court filing would constitute professional misconduct of the highest order — potentially warranting suspension or removal from the rolls.
• BCI Rule 9 and Rule 22: Rule 9 prohibits acting in a manner tending to mislead the court. Rule 22 requires an advocate who knows of a material error in a submission to promptly correct it. Mr. Parvaiz violated both — he submitted fabricated quotations (Rule 9) and then denied doing so rather than correcting them (Rule 22).
• Contempt of Courts Act 1971, Section 2(c): Wilful publication of false statements to the court that scandalise the administration of justice constitutes criminal contempt. Seven fabricated judicial quotations across a court submission, if wilfully made, would fall squarely within Section 2(c)’s scope.

Three Practical Tips for Indian Practitioners

1. Never deny AI use to a court when the evidence suggests otherwise. If a court identifies fabricated quotations and asks whether you used AI, and you did, say so — and take responsibility. The cover-up, as Justice Myers observed, is worse than the error. In India, dishonesty before the court is a distinct and more serious charge than inadvertent error.
2. The “I was called recently and made a mistake” defence has limits. Mr. Parvaiz noted he was called to the bar in 2022 and was a sole practitioner. Justice Myers did not accept that bar admission and ethics training could be completed without understanding that fabricating quotations is impermissible. Junior Indian advocates using AI tools face the same baseline standard.
3. Verify every quotation, not just every citation. Mr. Parvaiz’s case illustrates that checking the CanLII citation to confirm a case is real is not sufficient — you must also verify that the quoted passage appears in the judgment. Produce a screenshot or page reference from the original database showing the quoted text before including any quotation in a court submission.

Quick Takeaways

• Seven fabricated quotations from real cases — all with correct CanLII citations — is the most elaborate documented variant of AI hallucination in the current case law.
• Denying AI use when the evidence of fabricated quotations is overwhelming may constitute a separate, more serious dishonesty issue that transforms a professional error into potential misconduct.
• The Law Society of Ontario referral is the most severe judicial consequence yet documented in AI hallucination cases involving licensed practitioners — surpassing monetary sanctions and terminating dismissals (which affect parties, not lawyers).
• Ontario’s Rule 4.06.1(2.1) requires AI certification in court filings; failure to provide it is itself a procedural violation that the court can raise independently.
• In India: the BCI and State Bar Councils have equivalent jurisdiction to the Law Society of Ontario; AI-related misconduct referrals by Indian courts to the BCI are a near-term possibility as AI-impacted submissions increase.

Deep Dive: The “Cover-Up May Be Worse Than the Initial Error” — Lessons for AI Hallucination Disclosure

Kapahi v. Elite Real Estate represents the sharpest escalation in AI hallucination consequences yet recorded. The progression is instructive: (1) AI hallucinations in Reply Factum; (2) court confirms fabrication and inquires about AI use; (3) counsel denies AI use; (4) court finds denial implausible; (5) referral to Law Society; (6) potential criminal/disciplinary investigation. Each step added to the consequences.

Justice Myers’s observation that “perhaps doing it once could be some kind of slip or error that mistakenly found its way into the factum — but not seven times” articulates a principle that appears across AI hallucination cases: multiplicity of errors eliminates the innocent explanation. A single citation error might be carelessness. Seven fabricated quotations across a single Reply Factum, all attributing plausible-sounding legal principles to real cases that do not contain them, points to systematic AI generation.

The most important lesson for Indian advocates is the disclosure imperative. In every AI hallucination case where the lawyer acknowledged the error and disclosed it promptly — Mata v. Avianca, Fogarty v. Fogarty, even Burches v. Equifax where Burches admitted to fabrications before oral argument — courts responded with measured consequences: warnings, monetary sanctions, or admonishments. Where disclosure was denied or delayed, consequences escalated dramatically. Kapahi shows what happens at the extreme end of that spectrum.

For the Indian context, this is directly applicable to Bar Council disciplinary proceedings under Advocates Act Section 35. The BCI Disciplinary Committee considers aggravating factors including dishonesty, repetition of misconduct, and failure to remedy the situation. An advocate who discovers an AI error in their submission and discloses it immediately will face a very different disciplinary outcome than one who denies it under judicial inquiry. The disclosure calculus is clear: early, voluntary, complete disclosure is always the better path.