New York Appellate Sanctions for AI-Hallucinated Citations: A Nondelegable Duty to Verify Legal Authorities

1. Introduction

Deutsche Bank Natl. Trust Co. v LeTennier (2026 NY Slip Op 00040 [3d Dept Jan. 8, 2026]) arises from a long-running mortgage foreclosure action in Delaware County. Defendant Jean LeTennier executed a 2006 note secured by a mortgage and defaulted. Plaintiff Deutsche Bank National Trust Company, as Trustee, commenced foreclosure in 2018. After summary judgment for plaintiff was affirmed on appeal (189 AD3d 2022, 2025 [3d Dept 2020]), defendant continued filing repeated motions—many recycling standing arguments—prompting findings of frivolous conduct and “vexatious litigant” controls in Supreme Court.

On this appeal from multiple postjudgment orders, the substantive foreclosure issues were largely routine. The case became precedential because defense papers cited numerous cases that did not exist—conduct the Third Department identified as likely stemming from generative artificial intelligence (GenAI) “hallucinations.” The court used the appeal to articulate how New York’s sanctions framework applies when attorneys submit fabricated authorities and misstate real ones, and to emphasize that GenAI use does not diminish counsel’s duties of accuracy and candor.

2. Summary of the Opinion

Merits/foreclosure posture: The court affirmed Supreme Court’s denial of defendant’s CPLR 5015 motion to vacate because alleged “newly discovered” materials were publicly available or discoverable with diligence, and fraud allegations were conclusory.
Abandonment: Appeals from the August 2024 and September 2024 orders were deemed abandoned because defendant did not raise challenges to them in the opening brief.
GenAI misuse and fabricated citations: The court found at least 23 fabricated cases across five appellate filings, plus multiple misrepresentations of actual case law. Counsel conceded using AI in preparation.
Sanctions: Under 22 NYCRR 130-1.1, the court imposed (i) $5,000 on counsel for fabricated authorities, and (ii) additional sanctions for a frivolous appeal: $2,500 on counsel and $2,500 on defendant—totaling $7,500 against counsel and $2,500 against defendant.
Rule stated: Parties may use GenAI, but it never relieves an attorney of the obligation to fact-check and cite-check every filing; failure may be sanctionable depending on circumstances.

3. Analysis

A. Precedents Cited

i. CPLR 5015 relief: newly discovered evidence and fraud

The court grounded its CPLR 5015 analysis in established Third Department precedent:

HSBC Bank USA, N.A. v Sage, 143 AD3d 1214 (3d Dept 2016): quoted for the standard that relief may be granted for newly discovered evidence likely to change the outcome and not discoverable in time, or for fraud/misrepresentation/misconduct. The decision reinforced that CPLR 5015 is not a vehicle for re-litigating issues previously resolved.
Matter of Romine v New York Pub. Serv. Commn., 209 AD3d 1197 (3d Dept 2022), appeal dismissed 39 NY3d 1060 (2023): cited for the diligence requirement—new evidence must be shown undiscoverable earlier despite due diligence.
Wall St. Mtge. Bankers, Ltd. v Rodgers, 148 AD3d 1088 (2d Dept 2017): used to confirm the “could have been discovered” inquiry is fatal when the proffered materials were available earlier.
Carlson v Dorsey, 161 AD3d 1317 (3d Dept 2018), and Wells Fargo, N.A. v Levin, 101 AD3d 1519 (3d Dept 2012), lv dismissed 21 NY3d 887 (2013): invoked for the principle that conclusory fraud allegations unsupported by record evidence do not justify vacatur.

The court also relied on law of the case stemming from its own earlier foreclosure appeal (189 AD3d at 2024-2025), treating defendant’s renewed standing arguments as already decided and therefore not open for re-argument.

ii. Appellate abandonment

Matter of Shannon, 240 AD3d 1021 (3d Dept 2025), and Amici v Mazza, 234 AD3d 1170 (3d Dept 2025), lv denied 44 NY3d 902 (2025): cited for the settled appellate rule that issues (and entire appeals from specific orders) not argued in an opening brief are deemed abandoned.

iii. GenAI, hallucinations, and misstatements: persuasive authority and emerging case law

Because this was the Third Department’s first occasion to address GenAI misuse at the appellate level, it drew on a body of recent federal and state decisions confronting fabricated citations and AI-driven mischaracterizations:

Mata v Avianca, Inc., 678 F Supp 3d 443 (SD NY 2023), and Wadsworth v Walmart Inc., 348 FRD 489 (D Wyo 2025): cited for a balanced framing—GenAI can improve efficiency and access to justice, but hallucinations and citation fabrication pose serious risks, and submission of false authorities wastes adversarial and judicial resources.
United States v Google LLC, ___ F Supp 3d ___, 2025 WL 2523010, 2025 US Dist LEXIS 170459 (D DC Sept. 2, 2025): used to explain AI taxonomy and that “GenAI” is a subfield of AI; also cited in the broader discussion of hallucinations.
Sanders v United States, 176 Fed Cl 163 (Fed Cl 2025): cited for the practical problem that hallucinated cases may appear plausible (complete with reporter-like citations) but do not resolve to any real authority.
Green Bldg. Initiative, Inc. v Peacock, 350 FRD 289 (D Ore 2025), Kruglyak v Home Depot U.S.A., Inc., 774 F Supp 3d 767 (WD Va 2025), and Matter of Richburg, 671 BR 918 (Bankr D SC 2025): cited for the related risk that even when citations exist, GenAI may misstate holdings—often skewed in favor of the prompting party.
OTG New York, Inc. v OTTOGI Am., Inc., 2025 WL 2671460, 2025 US Dist LEXIS 183358 (D NJ Sept. 18, 2025): cited for the general proposition that filing hallucinated authorities may be sanctionable.
Ader v Ader, 87 Misc 3d 1213[A] (Sup Ct, NY County 2025), United States v McGee, ___ F Supp 3d ___, 2025 WL 2888065, 2025 US Dist LEXIS 201350, and Johnson v Dunn, 2025 WL 2086116, US Dist LEXIS 141805 (ND Ala July 23, 2025): cited for the important point that fabricated citations are sanctionable even if the “proposition” could have been supported by real law, because the misconduct lies in submitting false authorities and imposing verification burdens on courts and adversaries.

iv. New York sanctions doctrine: 22 NYCRR 130-1.1 and deterrence

WFE Ventures, Inc. v GBD Lake Placid, LLC, 197 AD3d 824 (3d Dept 2021): cited for New York courts’ discretion to impose costs or financial sanctions for frivolous conduct under 22 NYCRR 130-1.1.
Matter of Minister, Elders & Deacons of Refm. Prot. Dutch Church of City of N.Y. v 198 Broadway, 76 NY2d 411 (1990): cited for how courts assess “circumstances” in determining frivolousness and the systemic harm caused by misuse of judicial process.
Bell v State of New York, 96 NY2d 811 (2001): cited for sanction authority in frivolous and meritless appeals.
Engel v CBS, Inc., 93 NY2d 195 (1999): cited in the discussion of sanctions as a response to misuse of judicial process.
Levy v Carol Mgt. Corp., 260 AD2d 27 (1st Dept 1999): relied upon for the “goal oriented” nature of sanctions— punishment plus deterrence of the particular litigant and the bar at large.
De Ruzzio v De Ruzzio, 287 AD2d 896 (3d Dept 2001): used to characterize the appeal as a continuation of protracted, frivolous litigation pursued despite prior warnings and sanctions.

v. Sanctions range in AI-citation cases (comparators)

To calibrate proportionality, the court surveyed decisions imposing sanctions for fabricated AI authorities:

Cojom v Roblen, LLC, 2025 WL 3205930, 2025 US Dist LEXIS 225325 (D Conn Nov. 17, 2025)
Hall v Academy Charter Sch., 2025 WL 2256653, 2025 US Dist LEXIS 152628 (ED NY Aug. 7, 2025)
Lacey v State Farm Gen. Ins. Co., 2025 WL 1363069, 2025 US Dist LEXIS 90370 (CD Cal May 5, 2025)
Noland v Land of the Free, L.P., 114 Cal App 5th 426, 336 Cal Rptr 3d 897 (Cal Ct App 2025)
Mid Cent. Operating Engrs. Health & Welfare Fund v HoosierVac LLC, 2025 WL 1511211, 2025 US Dist LEXIS 100748 (SD Ind May 28, 2025)
Kruse v Karlen, 692 SW3d 43 (Mo Ct App 2024)

These cases supported the Third Department’s conclusion that meaningful monetary sanctions are sometimes warranted, especially where the misconduct is repeated, escalates after notice, or materially burdens the court.

B. Legal Reasoning

i. Disposition of the CPLR 5015 vacatur motion

The court treated defendant’s “new evidence” (SEC and Secretary of State-related materials) as a diligence problem rather than a merits problem: the evidence either predated summary judgment or was publicly accessible online. Under Matter of Romine v New York Pub. Serv. Commn. and HSBC Bank USA, N.A. v Sage, that failure ends the inquiry.

Fraud and misconduct allegations failed because they were conclusory and unsupported (Carlson v Dorsey; Wells Fargo, N.A. v Levin). The court also framed the motion as another attempt to repackage a standing challenge already decided in the 2020 appeal, invoking law-of-the-case finality.

ii. Abandonment as an appellate discipline tool

By deeming the August and September 2024 appeals abandoned (Matter of Shannon; Amici v Mazza), the court reinforced a practical lesson: even where a record is extensive and litigants are prolific motion filers, appellate review is limited to issues properly raised and argued.

iii. The core holding: fabricated authorities are per se frivolous under 22 NYCRR 130-1.1

The court’s key doctrinal move was to align GenAI hallucinations with existing New York sanctions categories—without needing a new rule. It held that submitting fabricated authorities is “completely without merit in law,” thus satisfying 22 NYCRR 130-1.1(c)(1), and also often implicates 22 NYCRR 130-1.1(c)(3) because it asserts materially false factual statements (i.e., that a case exists or holds X).

Importantly, the court rejected “it was just a citation error” as an adequate explanation, emphasizing the attorney certification duty (22 NYCRR 130-1.1a[b][1]) and the independent ethical duty of candor (Rules of Prof Conduct [22 NYCRR 1200.0] rule 3.3[a][1]). The opinion stresses that the obligation is not simply to avoid deliberate falsehoods, but to correct false statements and ensure that filings reflect a reasonable inquiry under the circumstances.

iv. Escalation after notice as an aggravating factor

The court treated counsel’s conduct as aggravated by repetition and continuation after notice. Defendant and counsel were alerted by plaintiff’s motion that cases were fabricated, yet subsequent filings added more fake cases and “false legal propositions.” Under 22 NYCRR 130-1.1(c), continuing conduct after its lack of basis “was apparent” weighs heavily toward sanctions.

v. Dual sanctions: (1) GenAI misuse and (2) frivolous appeal

The court separated two sanctionable categories:

Misuse of GenAI / fabricated authorities: $5,000 against counsel, reflecting the scale (23 fake cases), the continued use after notice, and the burden on the court.
Frivolous appeal as part of vexatious foreclosure resistance: relying on De Ruzzio v De Ruzzio, the court found the appeal itself to be a continuation of dilatory and harassing tactics, warranting an additional $2,500 against counsel and $2,500 against defendant.

The court also pointed to practical indicia that undermined counsel’s claimed oversight—such as metadata suggesting some documents originated from a program in the client’s name—supporting its conclusion that the appeal reflected a broader pattern of misconduct requiring deterrence.

C. Impact

i. A clear appellate rule for New York: GenAI is permitted, but verification is mandatory

This decision is positioned as New York’s first “appellate-level” sanctions opinion squarely addressing GenAI misuse. Its practical rule is straightforward: GenAI may be used as an assistive tool, but it does not reduce the lawyer’s nondelegable duty to verify citations, quotations, and propositions before filing.

ii. Likely downstream effects on motion practice and appellate briefing

More aggressive sanctions motions: Opposing parties who detect fabricated citations now have clear New York appellate authority to seek 22 NYCRR 130-1.1 relief.
Higher expectations of remedial action: The opinion treats “notice” as pivotal; counsel who promptly correct and disclose errors may fare better than those who “double down.”
Procedural hygiene: The abandonment holding underscores that prolific litigants cannot preserve issues by volume; they must brief them properly or lose them.

iii. Ethical and professional responsibility consequences

By anchoring its analysis in rule 3.3 (candor to the tribunal) and the certification obligation in 22 NYCRR 130-1.1a, the court implicitly frames GenAI citation hallucinations not merely as “sloppy lawyering,” but as conduct capable of raising professional-responsibility concerns when false statements are not corrected or when accuracy checks are not performed.

iv. Foreclosure litigation and vexatious tactics

Although the GenAI issue is broadly applicable, the setting matters: this was a foreclosure case with repeated, escalating postjudgment filings. The court’s willingness to sanction both counsel and client signals that New York appellate courts may use 22 NYCRR 130-1.1 to contain protracted foreclosure resistance where arguments are repetitious, previously decided (law of the case), and deployed to delay.

4. Complex Concepts Simplified

CPLR 5015: A rule allowing a court to relieve a party from an order or judgment in limited circumstances (e.g., truly “new” evidence or fraud). It is not a second chance to argue what was already decided.
“Newly discovered evidence” (in practice): Evidence must be genuinely unavailable earlier despite due diligence. If it was public, online, or obtainable with routine effort, it usually will not qualify.
Law of the case: Once an appellate court decides an issue (like standing), that determination generally governs later stages of the same case.
Abandonment on appeal: If an appellant does not argue an issue (or an order) in the opening brief, the appellate court treats it as waived/abandoned.
GenAI “hallucinations”: Outputs where a GenAI system generates plausible-sounding but false information—such as cases that do not exist or incorrect descriptions of holdings.
22 NYCRR 130-1.1 sanctions: New York’s rule permitting costs and financial sanctions for “frivolous conduct,” including meritless legal contentions or false factual statements.
Rule 3.3 (candor to the tribunal): Lawyers must not make false statements of law or fact to a court and must correct material false statements previously made.

5. Conclusion

Deutsche Bank Natl. Trust Co. v LeTennier affirms routine foreclosure-related orders but breaks new ground in New York appellate practice by imposing substantial sanctions for GenAI-driven fabricated case citations and misstatements of law. The Third Department’s central takeaway is that GenAI is not prohibited—yet it cannot be used as a substitute for professional judgment and verification. Submitting hallucinated authorities is, by definition, frivolous under 22 NYCRR 130-1.1, and persistence after notice significantly aggravates exposure. In parallel, the court confirms that appellate sanctions remain available to deter litigants and counsel from using repeated, previously rejected arguments to delay proceedings.