Unverified Generative-AI Case Citations Are “Frivolous Conduct” Under 22 NYCRR 130-1.1—Even for Pro Se Appellants

Case: Matter of Julien v Arthur, 2026 NY Slip Op 03308 (App Div 2d Dept May 27, 2026) (Wooten, J.)
Procedural posture: Father’s appeal from a Family Court custody order; Second Department issued an order to show cause regarding sanctions for fabricated authority in appellant’s brief.

I. Introduction

Matter of Julien v Arthur arises from related Family Court Act articles 6 (custody/parental access) and 8 (family offense/order of protection) proceedings between unmarried parents of a child born in June 2020. The Family Court granted the mother sole legal and residential custody, with defined parental access to the father. The father appealed pro se.

The appeal presented two distinct sets of issues:

• Merits: Whether the Family Court’s custody and parental access determinations were supported by a sound and substantial basis and served the child’s best interests.
• Appellate practice / integrity: Whether filing an appellate brief that includes a citation to a nonexistent case—apparently generated via generative artificial intelligence (GenAI) and not verified—constitutes “frivolous conduct” warranting sanctions under 22 NYCRR 130-1.1, even when the filer is pro se.

II. Summary of the Opinion

A. Sanctions holding

The Second Department found that the father’s brief cited a nonexistent Appellate Division, First Department decision as the sole authority supporting a claim of judicial bias, and the brief bore markers consistent with GenAI drafting (including bracketed “Appellant to provide…” placeholders). The Court concluded that the father’s failure to verify the citation—resulting in submission of fabricated legal authority—qualified as frivolous conduct under 22 NYCRR 130-1.1(c)(1) and (c)(3). The Court imposed a $250 monetary sanction and cautioned that future misuse could result in greater sanctions.

B. Disposition on the merits

On the custody and access issues, the Court affirmed. Applying the “best interests of the child” standard and deference to the Family Court’s credibility findings, the Second Department held the record supported awarding the mother sole legal and residential custody with certain parental access to the father. The father’s bias contention was also deemed unpreserved because he did not seek recusal.

III. Analysis

A. Precedents Cited

1. National GenAI “hallucination” cases used to frame the problem

The Court surveyed a growing body of federal decisions addressing GenAI-generated inaccuracies in legal filings, emphasizing that the technology can “hallucinate” plausible-looking but false citations, quotations, or statements of law.

• Smith v Smith; Jarrus v Governor of Michigan; OTG New York, Inc. v OTTOGI Am., Inc.; United States v McGee; Wadsworth v Walmart Inc.
  Cited for the general proposition that courts have recognized the pitfalls of unverified GenAI use in drafting briefs and memoranda.
• Braica v Frankowski
  Used to define and exemplify GenAI “hallucinations” and their forms (nonexistent cases, fabricated quotations, misstatements of law).
• Mata v Avianca, Inc. and OTG New York, Inc. v OTTOGI Am., Inc.
  Invoked for the institutional harms: time and expense imposed on adversaries to uncover false authorities and burdens on judicial resources.

2. Sanctions authorities for GenAI hallucinations (including appellate-level guidance)

• Fletcher v Experian Info. Solutions, Inc.; Amarsingh v Frontier Airlines, Inc.; Lexos Media IP, LLC v Overstock.com, Inc.; Dubinin v Papazian; Mata v Avianca, Inc.
  Collected to support the proposition that submitting GenAI-hallucinated content can constitute sanctionable frivolous conduct.

3. Pro se litigants and “baseline” obligations

• Button v McCawley; Kulkarni v Merit Sys. Protection Bd.; Moore v City of Del City; Anonymous v New York City Dept. of Educ.; Braica v Frankowski
  Cited to rebut the notion that pro se status immunizes a litigant from sanctions for fabricated citations; courts require a “baseline, common-sense” duty not to cite nonexistent law.

4. New York Appellate Department treatment of AI-related citation fabrication

• Dowlah v Professional Staff Congress (PSC-CUNY)
  First Department admonition: pro se status does not excuse failure to check legal citations; no monetary sanction imposed there.
• NewRez LLC v Morton
  First Department declined to sanction despite four nonexistent cases, while signaling that such citations “should generally warrant sanctions,” and that pro se status is not a categorical shield.
• Matter of Zareh
  First Department attorney discipline context: public censure for a GenAI-drafted brief containing citation errors and misrepresentations.
• Deutsche Bank Natl. Trust Co. v LeTennier
  Third Department imposed significant monetary sanctions for fabricated cases attributed to AI use “without human oversight,” and articulated factors for calibrating sanctions (volume of fake cases, continued misuse after notice, fake quotes, admissions/remorse, and impact on proceedings). This decision provided a direct New York template the Second Department adopted in principle while distinguishing on severity.

5. Sanctions framework under New York law

• Aretakis v CPEX Real Estate, LLC
  Quoted for the definition of frivolous conduct under 22 NYCRR 130-1.1(c) (no merit in law; delay/harassment; or false material factual statements).
• Grymes Dev. Co. v Fodera
  Cited for the proposition that even a single fictitious GenAI citation may warrant sanctions to account for wasted party/court resources.
• Leggio v Leggio
  Cited to emphasize deterrence as a core purpose of sanctions.

6. Custody/parental access standards and deference precedents

• Matter of Navarro v Clarke (best interests under totality of circumstances).
• Matter of Freyer v Macruari; Matter of Watling v Watling (deference to Family Court credibility and “sound and substantial basis”).
• Matter of Diaz v Vicente; Matter of Steward v Okon; Matter of Sookchan v Sookchan (stability, fostering relationship with other parent, and domestic violence/withholding the child as relevant best-interests factors).
• Matter of Calogero v McMahon (appellate restraint where credibility is pivotal).
• Matter of Krubally v Jobe (parental access schedule as discretionary determination requiring record support).

7. Preservation and bias/recusal precedents

• Matter of Baby Girl Z. [Yaroslava Z.]; Matter of George A.C. [Anthony C.]
  Stand for the preservation rule: a bias claim generally requires objection and a motion for recusal.
• Matter of Hossain v Chowdhury; Matter of Roman v Deceus; Matter of Perlman v Kolodny
  Explain the limited appellate inquiry into whether any bias unjustly affected the result, and illustrate the kind of record showing that defeats a bias claim (fair hearing, no predetermined outcome).

B. Legal Reasoning

1. The Court’s doctrinal move: GenAI misuse fits comfortably inside existing “frivolous conduct” categories

The Second Department did not create a new sanctions regime; rather, it treated GenAI-assisted fabrication as a modern pathway to traditional sanctionable misconduct. The father’s brief triggered two regulatory definitions:

• 22 NYCRR 130-1.1(c)(1): citing a nonexistent case as the sole support for a claim is “completely without merit in law” and not supportable as a good-faith extension or modification of law.
• 22 NYCRR 130-1.1(c)(3): submission of a fabricated citation constitutes asserting “material factual statements that are false” (i.e., the asserted existence and content of an authority).

2. The Court’s normative premise: verification is a non-delegable duty

The opinion’s core operational lesson is that GenAI is not a citator and not a substitute for basic validation. The Court framed the harm broadly—misleading courts and opponents, imposing costs to “expos[e] the deception,” and wasting judicial resources—making verification an essential condition of acceptable use.

3. Pro se status mitigates the amount, not the duty

The Court acknowledged that pro se litigants receive some leniency, and it distinguished the father’s conduct from Deutsche Bank Natl. Trust Co. v LeTennier based on:

• Scale (one nonexistent case vs. 23 fabricated cases in multiple filings),
• Post-notice conduct (no persistence after being alerted),
• Acceptance of responsibility and apology,
• Absence of counsel (pro se filer).

But those factors went to calibration, not immunity. Relying on Dowlah v Professional Staff Congress (PSC-CUNY), the Court made explicit that pro se status “does not excuse” failure to check citations and that sanctions also serve a deterrent function for other litigants and attorneys.

4. Tailoring the sanction

The Court selected a modest $250 sanction, justified as sufficient to:

• Address wasted institutional and adversarial resources,
• Mark the seriousness of fabricated authority,
• Deter recurrence—especially given the Court’s express warning of “greater” sanctions for future misuse.

5. Maintaining merits neutrality

Importantly, the Court stated it did not draw an unfavorable inference on the custody merits from the father’s GenAI misuse. This separation serves two functions:

• It confines punishment to the litigation misconduct (citation fabrication), not the substantive parenting claims.
• It protects the legitimacy of the best-interests determination by anchoring it to record evidence and credibility findings, not the briefing irregularities.

C. Impact

1. Second Department precedent on GenAI and sanctions

The decision squarely positions the Second Department alongside the Third Department’s more punitive approach in Deutsche Bank Natl. Trust Co. v LeTennier, while also acknowledging the First Department’s earlier cautionary/lenient responses in Dowlah v Professional Staff Congress (PSC-CUNY) and NewRez LLC v Morton. The net effect is to remove any remaining ambiguity within the Second Department: fabricated AI-generated authority is sanctionable, and pro se litigants are within the rule’s reach.

2. Practical consequences for appellate practice

• Verification becomes the focal compliance standard: The opinion effectively treats citation checking (e.g., confirming existence and holdings) as the minimum safeguard when GenAI is used.
• Orders to show cause may become more common: The Court’s sua sponte intervention via an order to show cause signals active policing of fabricated authority at the appellate level.
• Sanctions risk extends beyond attorneys: Pro se parties—often the most likely to rely on internet tools—are on notice that “good faith” reliance does not bar sanctions if the submission contains fabricated authority.

3. Likely doctrinal ripple effects

By mapping GenAI hallucinations onto 22 NYCRR 130-1.1(c)(1) and (c)(3), the Court provides a ready analytical framework for future cases involving:

• Fabricated quotations from real cases (a (c)(3) falsity and potentially (c)(1) meritlessness),
• Mischaracterizations of holdings (frequently (c)(3), sometimes (c)(1)),
• Repeated post-notice misuse (escalating sanctions, as emphasized in Deutsche Bank Natl. Trust Co. v LeTennier).

IV. Complex Concepts Simplified

• Generative artificial intelligence (GenAI): Software that produces text based on patterns in data. It can write plausible legal prose but may also generate incorrect “facts,” including fake cases.
• “Hallucinations”: A term courts use when GenAI outputs confident-sounding but false content—like a case that does not exist, a quote never written, or a rule a court never adopted.
• 22 NYCRR 130-1.1 (sanctions for frivolous conduct): A New York court rule allowing financial sanctions and costs when a party/attorney engages in frivolous conduct, including meritless positions or false material statements.
• “Material factual statements that are false” (22 NYCRR 130-1.1[c][3]): Not limited to historical facts; it includes representing that a legal authority exists and supports a proposition when it does not.
• Preservation / recusal motion (bias claims): Generally, to argue judicial bias on appeal, the party must object and move for the judge to recuse; otherwise the claim is often “unpreserved.”
• “Best interests of the child” and “sound and substantial basis”: Custody is decided based on the child’s welfare under all circumstances; appellate courts defer heavily to the trial court’s credibility assessments and will not disturb findings supported by the record.
• Writ for the child’s return: A court-directed mechanism used when a parent wrongfully retains a child contrary to an order, compelling return to the lawful custodian.

V. Conclusion

Matter of Julien v Arthur is a significant Second Department decision at the intersection of appellate integrity and emerging legal-technology risks. The Court held that submitting an appellate brief drafted with unverified GenAI output—where the brief includes fabricated legal authority—constitutes sanctionable frivolous conduct under 22 NYCRR 130-1.1, even for a pro se litigant. While the sanction here was modest ($250) due to mitigating factors (single fake citation, prompt acknowledgment, no persistence), the Court’s warning signals an escalation path for repeat or post-notice misconduct.

On the merits, the Court reaffirmed conventional custody principles: deference to Family Court credibility findings and a record-based best-interests analysis supporting sole custody to the mother with structured parental access to the father. The decision’s broader legacy, however, lies in its clear message that GenAI is permissible only with human verification—and that fabricated citations are not “mistakes” the appellate system will absorb without consequence.