Tribunal decision · Ontario CAT

Toronto Standard Condominium Corporation No. 2744 v. Meghan

2024 ONCAT 169 · November 18, 2024 · Partially upheld

TL;DR

ONCAT upheld a settlement-agreement-breach finding for food-delivery theft by the owner's guests (owner liable under s.119(2)) but refused to find a noise breach because decibel readings were taken outside the unit rather than inside the complainant's unit and the corporation failed to give the respondent any opportunity to address the noise before escalation. Notably, the tribunal flagged that the respondent's closing submission cited three AI-fabricated case citations — a direct caution against unverified AI-generated authority in enforcement filings.

Why this matters for Alberta owners

TL;DR: An owner defending against a settlement-breach allegation submitted three Ontario tribunal cases as authority in his closing submission. Member Spencer searched for each one. None existed. The Tribunal disregarded every legal argument Mr. Meghan built on those phantom cases — and ruled against him on the underlying dispute. This is the second published Ontario decision (after Vasina, 2026 ONCAT 3) catching AI-fabricated citations in an owner's filing, but it's the more instructive one for FineCheck users: unlike Vasina (records-only application), Meghan was a real fine-style dispute where the AI-generated arguments were supposed to be the owner's defense. The lesson for Alberta owners filing at the CDRT: a chatbot will confidently invent case names that look exactly like real ones, the tribunal WILL check, and your entire defense collapses when the cases don't exist.

The facts. Toronto Standard Condominium Corporation No. 2744 had a settlement agreement with David Meghan resolving an earlier set of food-theft and noise complaints. The corporation later alleged Mr. Meghan breached the settlement on multiple grounds, including two specific noise incidents on March 9 and July 13, 2024. The case went to a Stage 3 written hearing. In his closing submission — the moment in tribunal proceedings where parties cite legal authority for the propositions they want the tribunal to accept — Mr. Meghan included three citations to ONCAT decisions and used them to argue the tribunal should reject the corporation's noise allegations.

What the Tribunal did. From paragraph 29: "it appears that the citations may have been generated with AI assistance. A thorough search revealed that none of the cited cases exist." Member Spencer didn't just disregard the citations — she disregarded every finding Mr. Meghan attributed to those cases. The reasoning is structural: a legal argument is only as good as the authority it rests on. When the authority is fabricated, the argument is unsupported, and an unsupported argument loses on its merits. Mr. Meghan then lost on the substantive issues too: the Tribunal found he had breached the settlement agreement on the food-delivery incidents (his guests had taken other residents' food deliveries from the lobby) and ordered $125 in tribunal-fee reimbursement.

Why this matters more than Vasina. We've already covered the AI-citations problem in the context of Vasina v. YCC 486 (see /tribunal-decisions/2026-oncat-3) — where a self-represented owner used AI to construct a records-dispute application and was caught with four phantom cases. Meghan is the same failure mode, but in the context most Alberta owners will actually face: defending against an allegation of bylaw breach. Vasina shows the offensive failure mode (filing an application built on phantom cases). Meghan shows the defensive failure mode (defending against an allegation with phantom cases). For the typical FineCheck user — an owner who received a fine notice and wants to challenge it — the defensive failure mode is the one that matters. You're not filing a records dispute. You're trying to keep a fine from sticking. If your defense relies on cases the AI invented, you lose the defense.

The other lesson from Meghan: decibel readings taken outside your unit don't actually prove you were too loud. This is the second important holding from Meghan, and it's the one Alberta owners should know about most directly. The corporation here claimed two breach-of-settlement noise incidents. On the July 13, 2024 incident, security staff took decibel readings outside Mr. Meghan's door. Those readings appeared to be high. Member Spencer rejected them anyway, holding at paragraph 33: "decibel readings taken in the hallway outside a unit's door only serve to confirm a sound's source; they are not indicative of the noise level a complainant is experiencing in their own unit." The Tribunal also noted that Mr. Meghan had not been contacted and given an opportunity to address the noise before the corporation escalated. Both defects fell on the side of the corporation, and Member Spencer found insufficient evidence of unreasonable noise on either incident — both noise findings dismissed.

Why this matters for fine notices. A fine notice that rests entirely on hallway-readings evidence — something like 'we measured 65 decibels outside your door, that's loud, here's your fine' — has the same fundamental defect as Meghan: hallway readings prove a sound came from your unit, but they don't prove the sound was unreasonable to the complainant in their unit. For an Alberta fine notice to substantially comply with CPR s.73.7's contravention-fact requirements, the evidence underlying the alleged breach must actually establish the breach. Decibel readings taken in the wrong place don't do that. When FineCheck flags a notice as having insufficient evidentiary support for the alleged breach, this is the kind of authority that backs the flag up.

Action steps for owners

What to actually document if you're the owner. The Meghan analysis tells you what kind of evidence the tribunal will accept and what it won't. If you're alleged to be the noise source, your defensive evidence should focus on what was happening in your unit: the activities the corporation calls "unreasonable" measured against ordinary-living standards, the times of day involved, whether you were even home, and whether the corporation gave you a chance to address the issue before escalating. Tribunal members will not accept a corporation's word that hallway readings mean what the corporation says they mean — they require objective evidence of the disturbance in the complainant's unit, and they require the corporation to have given the alleged offender a chance to address the complaint. If those two things are missing from the fine notice's evidentiary basis, your CDRT submission should say so plainly.

Why chatbots do this. Large language models trained before late 2023 are particularly prone to fabricating tribunal citations because their training data underweights specialized administrative tribunals like ONCAT or the Alberta CDRT. The model has seen the FORM of ONCAT citations — a year, the abbreviation, a number — and will produce something that looks correct (think 'Smith v. TSCC 1234, 2022 ONCAT 89') even when no such case exists. Asking the chatbot to verify its own citations does not work; it will confidently confirm fabricated cases as real. The structural fix is to use a tool that pegs every citation to a verified source. That's not a vague promise about AI safety — it's the actual product behaviour. FineCheck's reports cite the Condominium Property Regulation by section number, every citation is pegged to verified statute text in the codebase, and the report makes the verification path explicit. When the AI is uncertain, the report says so rather than confabulating.

Action steps for owners

What to do if you've already filed something built on AI-generated cases. Withdraw the citations before the tribunal flags them. Member Spencer's response in Meghan was relatively mild — she disregarded the arguments and ruled on the rest of the record. The Vasina decision suggests that response may not last. Tribunals across Canada are now actively watching for AI-fabricated authorities; treating the same failure as merely "disregard" rather than as bad faith was Member Spencer's discretion in 2024. By late 2026, that discretion is likely to harden, and tribunals may start treating AI-fabricated submissions as conduct warranting costs orders. If you submitted citations you can't independently verify on CanLII, file a correction with the tribunal before they catch it.

The bottom line: Meghan is the canonical example of why AI tools without verified citation grounding are dangerous in a tribunal context, especially when you're the defendant. The Alberta CDRT has been operating since April 1, 2026 and will follow the same pattern: cases get checked, fabrications get disregarded, the underlying argument collapses with them. FineCheck's verified-citation framework exists precisely so the AI never gets the chance to invent. Run FineCheck on the fine notice itself, build your CDRT submission on the verified statutory and precedent citations the report flags, and verify every additional case you cite by opening it on CanLII before you submit.

What the tribunal said

Selected excerpts from the Ontario CAT's reasoning. Full decision on CanLII.

[29] Mr. Meghan denies that he made unreasonable noise on either March 9, 2024 or July 13, 2024. In support of his position, he referred me to three Tribunal decisions in his closing submission. While he included full case citations, it appears that the citations may have been generated with AI assistance. A thorough search revealed that none of the cited cases exist. Further, they are dated before January 1, 2022, the date when the Tribunal's jurisdiction expanded to include nuisance disputes. Therefore, I have not considered any of these decisions' findings to which Mr. Meghan referred. [33] The frequency and duration of noise and the time of day at which it occurs are all considerations in determining whether noise is unreasonable. ... However, in this case, there was only one complainant in each of the two incidents the corporation alleges were a breach of the Agreement. In such cases, decibel readings may be a helpful determinant. However, decibel readings taken in the hallway outside a unit's door only serve to confirm a sound's source; they are not indicative of the noise level a complainant is experiencing in their own unit. [35] On July 13, 2024, while security staff did take decibel readings and those readings do appear to be high, they were taken outside Mr. Meghan's door. Again, there is no evidence that staff assessed the noise level in the complainant's unit. Nor is there any evidence that Mr. Meghan was contacted and given an opportunity to address it.

Connects to FineCheck's framework

Got a similar Alberta condo fine? A FineCheck report applies the same substantial-compliance + prejudice framework the tribunal used here. $15. Run a check on your notice →

Jurisdictional note: Ontario CATdecisions are not binding on Alberta's CDRT, but the CDRT will look to other Canadian condo tribunals for guidance under analogous statutory provisions until its own case law develops. The substantial-compliance + prejudice framework used in this decision parallels the test in CDRT Policies and Procedures s.7(b).

Decision date: November 18, 2024 · Citation: 2024 ONCAT 169 · Outcome: partially upheld. FineCheck's commentary is published for research and educational use; not legal advice. Verify any reliance on this decision against the original text.