Sanctions for GenAI-Hallucinated Authority: Lawyers Must Cite-Check and Remain Candid to the Court
1. Introduction
Landberg v City of New York (2026 NY Slip Op 03935 [2d Dept June 23, 2026]) is an Appellate Division,
Second Department decision arising not from the merits of a sidewalk trip-and-fall appeal, but from the integrity of the
appellate process itself. The case began as a personal injury action in which plaintiff Judith Landberg alleged she was
injured when she tripped on a loose brick in a tree well abutting property owned by defendant Whitman Owner Corp.,
with the City of New York also named as a defendant.
The Supreme Court, Kings County, granted summary judgment to both defendants, dismissing the complaint. The Second Department
affirmed that order on the merits in a separate decision (Landberg v City of New York, 2026 NY Slip Op 03433).
The present opinion is a Decision and Order on Motion issued after the court, on its own motion, required
counsel to show cause why sanctions should not be imposed under 22 NYCRR 130-1.1 for filing an appellate brief
containing fabricated and erroneous legal authority—conduct the court found was linked to unverified use of generative
artificial intelligence (“GenAI”).
The core issues were: (1) whether filing a brief containing nonexistent cases, fabricated quotations, and mischaracterizations
of real precedents constitutes “frivolous conduct” under 22 NYCRR 130-1.1; (2) how an attorney’s duties of competence, candor,
and supervision apply when GenAI is used in legal drafting; and (3) what sanctions are appropriate to punish and deter such
conduct.
2. Summary of the Opinion
The Second Department imposed monetary sanctions on both plaintiff’s appellate counsel and his law firm after determining that
the appellant’s brief—prepared with the assistance of GenAI—contained:
- citations to nonexistent cases;
- fabricated purported Court of Appeals quotations “wholly contrary to actual law”;
- and misrepresentations of what real cases held.
The court sanctioned Michael Sanders, Esq. in the amount of $8,000 and sanctioned
Law Offices of Michael S. Lamonsoff, PLLC in the amount of $2,500, payable to the
Lawyers’ Fund for Client Protection of the State of New York. The court declined to award costs because the
defendants did not seek them and submitted no billing records. The court also noted that separate attorney disciplinary action
“may be a matter for the Attorney Grievance Committee.”
3. Analysis
A. Precedents Cited
1) 22 NYCRR 130-1.1 as the governing sanctions framework
The opinion grounds the sanction power in 22 NYCRR 130-1.1, which authorizes costs and sanctions for
“frivolous conduct,” including conduct that is (1) completely without merit, (2) undertaken to delay/harass, or (3) asserts
material factual statements that are false. The court emphasizes the rule’s requirement that the tribunal consider the
circumstances, including time available to investigate and whether counsel persisted after the lack of basis should have
been apparent.
2) Deutsche Bank Natl. Trust Co. v LeTennier
The court relies on Deutsche Bank Natl. Trust Co. v LeTennier (2026 NY Slip Op 00040) for two propositions:
(i) sanctions serve both retributive and deterrent purposes; and (ii) attorneys are not
barred from using GenAI, but GenAI use does not abrogate the obligation to fact-check and cite-check “every
document filed with a court.” This case supplies the court’s policy rationale for treating AI-related errors not as a
novelty excuse but as a compliance problem measured against longstanding professional duties.
3) Matter of Julien v Arthur
Matter of Julien v Arthur (2026 NY Slip Op 03308) is the Second Department’s recent, directly analogous
GenAI-sanctions discussion. There, a pro se litigant cited a nonexistent case as the sole support for a claim of bias; the
court sanctioned him $250. Landberg uses Julien as a baseline comparator: pro se status, single fake case,
and prompt acknowledgment justified a low sanction in Julien, whereas a trained attorney’s multi-layered
misrepresentations and lack of candor justified a substantially higher sanction here.
4) Mata v Avianca, Inc.
The opinion quotes Mata v Avianca, Inc. (678 F Supp 3d 443 [SD NY]) for an influential articulation of the
harms caused by fake authorities: wasted resources, prejudice to clients, reputational harm to judges and parties, and
broader cynicism about the profession and judicial system. Mata functions as an external validation of the
institutional stakes—i.e., the injury is not merely an “oops” in a brief but an attack on the reliability of adjudication.
5) Grymes Dev. Co. v Fodera
Citing Grymes Dev. Co. v Fodera (88 Misc 3d 767 [Sup Ct, Richmond County]), the court links AI misuse to the
lawyer’s ethical obligation to provide independent professional judgment. The court uses this to reinforce
that delegating thinking to GenAI without verification is not merely sloppy research; it can be inconsistent with the
attorney’s role as a fiduciary decision-maker responsible for the content presented to courts.
6) Other GenAI-sanctions authorities used to support the sanction level
The court also cites Idehen v Stoute-Philip (2025 NY Slip Op 51211[U] [Civ Ct, Queens County]) and
Fletcher v Experian Info. Solutions, Inc. (168 F4th 231 [5th Cir]) among authorities reflecting a growing
body of case law treating AI-generated falsehoods as sanctionable when filed without adequate human oversight.
7) Matter of Zareh (discipline context)
The decision cites Matter of Zareh (2026 NY Slip Op 00619) to show that GenAI-related citation errors and
misrepresentations can have consequences beyond motion practice—namely, attorney discipline. The reference
helps frame the court’s closing observation that grievance review may be appropriate.
8) Merits-law precedents invoked (and misused) in the brief
A central feature of the opinion is that it details exactly how the appellant’s brief corrupted the legal record:
-
The brief invented a Court of Appeals case, Xiang Fu Ji v City of New York (13 NY3d 791, 794), including a
fabricated quote asserting Administrative Code § 7-210 should be “construed liberally.”
-
The court contrasted that falsehood with the real governing Court of Appeals rule in
Vucetovic v Epsom Downs, Inc. (10 NY3d 517, 521): because § 7-210 is “in derogation of common law” and
creates new liability, it “must be strictly construed.”
-
The court acknowledged the existence of Xiang Fu He v Troon Mgt., Inc. (34 NY3d 167), a real case
analyzing § 7-210, but noted it did not contain the claimed “remedial statute” language.
-
The brief also invented a quotation from a fictitious Hausser v Giunta (88 AD3d 969, 970). The court noted
that while a real Hausser v Giunta (217 AD2d 604, revd 88 NY2d 449) exists, it concerned a different
municipality’s code and did not contain the asserted language.
-
The brief cited a wholly nonexistent Lack v Lack (841 NYS2d 403, 404) for the de novo standard of review.
-
The brief misrepresented Campaign for Fiscal Equity v State of New York (86 NY2d 307, 318) as addressing
strict construction of exceptions to remedial statutes, when the cited passage in fact concerned CPLR 3211(a)(7) pleading
sufficiency.
-
The brief mischaracterized Rodgers v City of New York (34 AD3d 555) as holding that a “two-year search” can
be sufficient to establish lack of prior written notice, when Rodgers did not address prior written notice or
municipal record searches.
These merits-law citations are important not because the court revisits the tort issues (it does not), but because they show
the depth of the problem: the brief did not merely include clerical citation mistakes; it introduced fabricated authority and
distorted real holdings in ways that could mislead a court on core interpretive rules (e.g., strict vs liberal construction).
B. Legal Reasoning
The court’s reasoning proceeds in a structured sanctions analysis under 22 NYCRR 130-1.1:
-
Identification of false “material factual statements”: A case citation and quotation presented as legal
authority is treated as an asserted fact about what the law is and what courts have said. Nonexistent cases and invented
quotations therefore fall squarely within “material factual statements that are false” (22 NYCRR 130-1.1[c][3]).
-
Assessment of culpability and surrounding circumstances: The court contrasted a pro se litigant’s mistake
in Matter of Julien v Arthur with a licensed attorney’s conduct here. It stressed Sanders had access to Westlaw and
is bound by professional rules, and that the unreliability of GenAI (including “hallucinations”) has been widely publicized
for years, putting the bar on notice.
-
Aggravating factor—lack of candor at oral argument: When questioned, Sanders declined a recess and
represented that the authorities came from Lexis/Westlaw/books/other briefs; he later conceded he “should have stated that
they were hallucinated by artificial intelligence” and admitted fear motivated his failure to be candid. The court treated
this not as a minor misstep, but as demonstrating “professional immaturity, arrogance, and profound lack of respect for the
judicial system.”
-
Aggravating factor—scope and nature of the misinformation: The court found it implausible that GenAI was
limited to “supplemental research,” because the brief also contained fabricated quotations and reversals of actual law
(notably the strict-construction rule from Vucetovic v Epsom Downs, Inc.), plus misrepresentations of real cases.
This broadened the inference of inadequate supervision and verification.
-
Sanction calibration—punishment and deterrence: Drawing on Deutsche Bank Natl. Trust Co. v LeTennier
and the cited GenAI-sanctions cases, the court selected monetary sanctions aimed at deterring both the attorney and the bar
at large from filing unverified AI-generated content.
-
Law firm accountability: Although the firm presented evidence that Sanders exceeded firm policy, the court
still sanctioned the firm because its name appeared on the brief and deterrence warranted institutional responsibility.
This reflects a principle of professional practice: courts often treat filings as the product of the attorney and the firm,
and internal policy is not a complete shield when the court’s processes have been burdened.
C. Impact
1) A clear New York appellate warning on GenAI use: The opinion reinforces an emerging rule in New York:
attorneys may use GenAI, but they must implement human verification equivalent to supervising work product
from a subordinate (paralegal/intern/other attorney), as framed in Deutsche Bank Natl. Trust Co. v LeTennier.
Landberg makes the sanction exposure concrete and public.
2) Sanction levels may scale with professional status and candor: By contrasting the $250 sanction in
Matter of Julien v Arthur with $8,000 here, the Second Department signals that courts may treat:
(i) licensed counsel as more culpable than pro se litigants,
(ii) multiple fabricated authorities as more egregious than a single error, and
(iii) lack of candor as a major aggravator.
3) Institutional responsibility for firms: The separate $2,500 sanction against the firm underscores that AI
governance is not only an individual ethics issue; it is also a law-office systems issue. Firms can expect
scrutiny of training, approved tools, supervision, and enforcement if a filing under the firm’s name contains fabricated
authority.
4) Possible ripple effects on motion practice and appellate briefing: Litigants may increasingly request
sanctions, demand disclosure of AI use, or insist on certification of citation checking. Even absent such formal requirements,
Landberg may change norms: courts and adversaries will be more willing to investigate suspicious citations, and
appellate counsel will be expected to have immediate answers when questioned about sources.
5) Ethical and disciplinary exposure: By referencing Matter of Zareh and explicitly pointing to the
Attorney Grievance Committee, the court highlights that GenAI-related misstatements can move from the realm of sanctions into
professional discipline where misrepresentations implicate rules of candor and competence.
6) A notable side note on adversary conduct: The opinion remarks that although the City and the owner filed
respondents’ briefs, neither alerted the court to the fabricated citations and quotations. While the court does not sanction
them, the observation implicitly reinforces that all “officers of the court” share an interest in protecting the integrity of
the decisional process—particularly where fabricated authority may mislead.
4. Complex Concepts Simplified
- “GenAI hallucinations”
-
Outputs that look authoritative (case names, citations, quotations) but are false—e.g., a made-up case, an invented quote, or
a claim about what a real case held that is untrue.
- 22 NYCRR 130-1.1 sanctions vs “costs”
-
The rule allows a court to impose (a) costs (reimbursement of actual expenses/attorney’s fees caused by frivolous
conduct) and/or (b) a financial sanction (a penalty). Here, the court imposed sanctions but declined costs because
defendants did not seek them and provided no billing records.
- “Material factual statements that are false” in a legal brief
-
A brief’s representations about what cases exist and what they say are treated as “facts” about legal authority. When those
representations are fabricated, they can qualify as materially false statements under the sanctions rule.
- Strict vs liberal construction (Administrative Code § 7-210)
-
If a statute changes the common law by creating liability where none existed, courts often say it must be strictly
construed—applied narrowly and according to its text. The opinion highlights that the appellant’s brief falsely claimed
the opposite (liberal construction) and falsely attributed that view to the Court of Appeals, contrary to
Vucetovic v Epsom Downs, Inc..
- “Prior written notice” (Administrative Code § 7-201[c])
-
Many municipal liability regimes require proof the City received written notice of a defect before it can be held liable,
subject to exceptions. The opinion’s focus is not the doctrine itself, but that the brief misrepresented
Rodgers v City of New York as speaking to how long the City must search its records to show lack of notice.
- Duty of candor
-
Lawyers must be truthful to the court, including when questioned about sources. Here, the court treated counsel’s
non-disclosure/misleading explanation at oral argument as a serious aggravating factor.
5. Conclusion
Landberg v City of New York establishes a pointed New York appellate precedent on GenAI in litigation: while
GenAI tools are not prohibited, attorneys and firms remain fully responsible for verifying every citation, quotation, and
characterization of authority submitted to a court. Filing GenAI-assisted work product containing nonexistent cases, fabricated
quotations, and distorted holdings constitutes sanctionable frivolous conduct under 22 NYCRR 130-1.1,
particularly where the attorney fails to act with candor when confronted.
The decision’s significance lies in its practical message to the bar: GenAI can be used, but unchecked GenAI cannot be
filed. The court’s dual sanctions—$8,000 against counsel and $2,500 against the firm—reflect a deterrence-centered view
that responsibility for AI governance is both individual and institutional, and that the integrity of legal authority in briefs
is foundational to the judicial system.