Executive Summary: Key Legal and Evidentiary Issues

• Judicial concern centered on seven fabricated “quotations” in a reply factum, each falsely attributed to real reported decisions.
• The court grappled with whether these fake passages resulted from undisclosed use of generative AI or from deliberate/human fabrication, and how to respond without an investigative fact-finding process.
• Prior proceedings involved enforcement of an arbitral costs award and an unsuccessful Rule 59.06 motion to vary Steele J.’s order, raising issues of proper use of post-judgment procedural remedies.
• A contemplated personal costs motion under Rule 57.07 highlighted potential sanctions against counsel for litigation misconduct tied to the false authorities.
• The matter engaged professional responsibility and legal ethics, including failure to provide the required AI-use certification and the integrity of representations made to the court.
• Ultimately, the judge declined to launch contempt proceedings and instead referred the decision to the Law Society of Ontario, leaving any further investigation and discipline to external authorities.

Background and procedural history

The dispute arises out of earlier arbitral and court proceedings between Kapahi Real Estate Inc. and related parties as Applicants, and Elite Real Estate Club of Toronto Inc. and others as Respondents. An arbitral tribunal had held the Respondents liable in costs, and the Applicants then brought an application to enforce that arbitral costs decision. Steele J. of the Ontario Superior Court of Justice enforced the tribunal’s decision and also awarded costs of that enforcement application against the Respondents. The exact dollar amounts of the arbitral costs and of Steele J.’s costs award are not specified in this endorsement.
The Respondents subsequently brought a motion under Rule 59.06 of the Rules of Civil Procedure seeking to vary Steele J.’s order. They argued, first, that there had been a “slip” such that the formal order did not reflect the intention expressed in her endorsement, and, second, that new facts had emerged undermining her decision. That motion came before Myers J. and was “wholly unsuccessful,” leaving Steele J.’s enforcement and costs order intact. In his January 7, 2026 decision dismissing the motion to vary (reported at 2026 ONSC 132), Myers J. also rejected allegations by Respondents’ counsel, Mr. Khalid Parvaiz, that Applicants’ counsel had engaged in “sharp practice,” finding those accusations scurrilous and unsupported.

The reply factum and the alleged AI hallucinations

The present endorsement was prompted by concerns about the integrity of authorities cited in the Respondents’ reply factum on the motion to vary, prepared and signed by Mr. Parvaiz. After the dismissal of the Rule 59.06 motion, the Applicants filed written costs submissions seeking elevated costs on the basis of the Respondents’ conduct. Among other complaints, they highlighted what appeared to be AI-type hallucinations in the Reply Factum: passages that purported to be direct quotations from binding or persuasive case law but did not actually exist in the decisions cited. The Applicants also notified the court of their intention to bring a motion for costs against Mr. Parvaiz personally under Rule 57.07, which allows for costs orders against lawyers for undue delay, negligence, or other misconduct in litigation.
In response, Myers J. independently checked the cases referenced in the impugned passages and confirmed that the Reply Factum contained quotations that did not originate from the decisions to which they were attributed. Through the court registrar, he advised counsel by email that he had verified the misquotations, asked whether generative AI had been used to draft the factum, and queried why the Reply Factum lacked the certification required by Rule 4.06.1(2.1) regarding the use of generative AI or similar tools in preparing court documents.

Settlement of costs and abandonment of the personal costs motion

As these concerns unfolded, Mr. Gavin Tighe was retained to act for Mr. Parvaiz personally. He asked the court to defer deadlines for the Respondents’ costs submissions pending a case conference to address scheduling of the proposed personal costs motion under Rule 57.07. Case conferences were held to deal with those issues. During the first conference, Mr. Tighe indicated that his client would respond in writing to the judge’s questions about the Reply Factum and AI.
Before any formal Rule 57.07 motion materials were filed and adjudicated, the parties reached a settlement regarding costs of the unsuccessful motion to vary. By letter of February 26, 2026, Mr. Tighe advised that the Respondents had agreed to pay the Applicants’ costs of that motion on a substantial indemnity basis fixed at $32,747.40, and that the Applicants were abandoning their motion for costs against Mr. Parvaiz personally. A consent order implementing this settlement was signed by Myers J. The endorsement notes that there was no disclosure—nor any obligation to disclose—whether Mr. Parvaiz or his insurer contributed personally to that payment; therefore, the court could not determine whether any part of the $32,747.40 might already represent a form of consequence for the misquoted authorities.

The seven fabricated quotations

Despite the settlement of monetary costs, the court remained concerned about the integrity of the material filed and the explanation for how seven false “quotations” came to be placed in the Reply Factum. The endorsement reproduces and analyzes those seven passages.
In each instance, the Reply Factum set out text inside quotation marks, usually indented, and cited real decisions with correct neutral citations. The cases were genuine, and the neutral citations (often to CanLII) were accurate, but the quoted words were not. Crucially, the footnotes did not provide pinpoint paragraph references, which would normally show where in the decision the quotation is found. Instead, the Reply Factum presented fully formed, doctrinally coherent statements that supported the Respondents’ arguments—but which did not exist in the authorities cited.
Examples included a supposed quotation from Sattva Capital Corp. v. Creston Moly Corp., 2014 SCC 53, describing arbitration as a “creature of contract” and articulating a narrow, deferential role for courts, culminating in the statement that “the arbitrator is not required to address issues that are not before them or to go beyond the scope of what the parties have asked them to decide.” Nothing remotely like this appears in the actual Sattva decision. Similar made-up doctrinal statements were attributed to Ontario Court of Appeal cases on piercing the corporate veil and personal liability of directors and officers, such as FNF Enterprises Inc. v. Wag and Train Inc., Pita Royale Inc. (Aroma Taste of the Middle East) v. Buckingham Properties Inc., and BH Frontier Solutions Inc. v. 11054660 Canada Inc. (Canadian Choice Supply). Again, the cases exist, but the quoted passages do not.
Two separate fabricated quotations were also attributed to Trustees of the Millwright Regional Council of Ontario Pension Trust Fund v. Celestica Inc. In those instances, the factum ascribed to “Justice Mosley” a formulation of Rule 59.06(1) that extended the rule beyond simple typographical errors to “omissions of substance where the court failed to address a matter that should have been addressed.” In reality, the decision was written by Perell J., not Mosley J., and while there is some resemblance between the first part of one fabricated quotation and an actual paragraph of the decision, the rest of the text is imaginary. The overall pattern led the court to conclude that these were not minor misquotations or transcription slips; they were entirely invented doctrinal statements dressed up as direct extracts from case law.

Counsel’s explanation and the court’s concerns

After the costs settlement, Mr. Parvaiz wrote directly to the court, addressing the judge’s questions. He acknowledged that his recitation of the legal principles and substance of the cases was not accurate, identifying several paragraphs of his Reply Factum—though he undercounted them—as misrepresenting the authorities. He further admitted that he had failed to include the required Rule 4.06.1(2.1) certification concerning AI use. He characterized both the misstatements of law and the missing certification as “clear errors” caused by “a lack of due care,” “misreading of the cases cited,” “carelessness,” and “inadvertence.” He expressly denied using or relying on artificial intelligence or similar tools in preparing the Reply Factum, asserted that the errors were unintentional, took full responsibility, and noted that he is a sole practitioner called to the Ontario bar in 2022. He emphasized that he had reflected on the mistakes, understood the court’s need to rely on counsel, and deeply regretted what had occurred.
Myers J. accepted that the cases cited were real and that the neutral citations were correct, but he found the explanation difficult to reconcile with the nature of the errors. If no AI was used, the judge queried how a lawyer could “misread” a case into generating seven fully formed paragraphs purporting to be quotations but actually invented. He questioned how mere carelessness or inadvertence could explain fabricating language that squarely supported counsel’s argument and then enclosing it in quotation marks as precedent. While one erroneous “quotation” might conceivably be a stray blunder, repeating this pattern seven times suggests either that AI-generated hallucinations were pasted into the factum without adequate verification, or that counsel believed he could manufacture legal principles and attribute them to existing authorities. Either scenario raised serious concerns about honesty and professional conduct.
The judge also considered the possibility that if AI had been used and not disclosed, the subsequent denial of AI use could amount to a “cover-up” worse than the initial misconduct. He referenced his prior decisions in Ko v. Li, where he had required counsel to show cause why she should not be held in contempt for fake authorities and ultimately referred the matter to the Attorney General after the lawyer admitted untruthfulness. Here, by contrast, there was no such admission, and the role of AI, if any, remained unclear.

Judicial options and the chosen response

Confronted with these uncertainties, Myers J. evaluated the appropriate judicial response. In a contempt-style show-cause hearing, a judge lacks investigative tools such as access to computer metadata, browser histories, or the ability to cross-examine witnesses in a prosecutorial manner. At the same time, a Rule 57.07 personal costs motion related to these same facts had already been threatened and then effectively resolved when the Respondents agreed to pay substantial indemnity costs of $32,747.40 for the failed motion to vary, with any personal exposure of counsel potentially addressed in that settlement.
The judge acknowledged that contempt and other quasi-criminal or criminal processes, as well as regulatory action by the Law Society of Ontario, remain available and are not precluded by the earlier costs process. Nevertheless, he considered that the existence of the prior Rule 57.07 jeopardy and the limits of the court’s fact-finding capabilities were reasons to temper further direct judicial intervention. There are, he observed, other institutions—such as the Toronto Police Service and, especially, the Law Society of Ontario—better equipped to investigate whether the fake quotations resulted from undisclosed AI use, deliberate fabrication, or some other form of professional misconduct.
Accordingly, rather than initiating contempt proceedings or attempting to conduct his own de facto investigation, Myers J. concluded that the proper course was to leave any deeper inquiry to the relevant authorities. He determined that the best mechanism for ensuring accountability and clarifying what occurred was to refer the decision to the Law Society of Ontario. The endorsement formally records his intention to do so.

Outcome and implications

The practical outcome of these related proceedings can be summarized on two levels: the underlying litigation and the professional-responsibility dimension. On the litigation front, the Respondents’ attempt to vary Steele J.’s enforcement and costs order under Rule 59.06 failed completely, leaving the Applicants’ arbitral award and prior costs victory in place. As part of a negotiated resolution of the post-judgment costs dispute, the Respondents agreed to pay the Applicants substantial indemnity costs of $32,747.40 for the unsuccessful motion to vary. The exact total of all awards flowing from the arbitration, Steele J.’s order, and subsequent costs rulings cannot be determined from this endorsement alone, because the earlier cost figures are not specified.
On the professional-conduct front, the court exposed and analyzed a disturbing pattern of seven fabricated quotations, all presented as authoritative legal statements from real decisions. While the judge remained unable to determine definitively whether these passages originated from AI hallucinations or from human invention, he rejected the suggestion that they could be explained by mere misreading or inadvertence. Weighing the limits of a contempt-style inquiry against the existence of specialized investigative bodies, he opted not to pursue contempt or further in-court sanctions at this time. Instead, he referred the matter to the Law Society of Ontario for its consideration, leaving it to the regulator (and, potentially, other authorities) to decide whether to investigate or lay charges. In this sense, the Applicants emerge as the successful party in the underlying motion-to-vary litigation and its costs sequel, with at least $32,747.40 ordered in their favour on a substantial indemnity basis, while the full aggregate of all costs and monetary consequences across the arbitration and court proceedings cannot be precisely quantified on the basis of this decision alone.