Submitting Briefs with Nonexistent Case Citations—Whether AI-Generated or Not—Constitutes Prejudicial Conduct Under D.C. Rule 8.4(d) and Warrants Fourth Circuit Discipline

Case: In re: Eric Chibueze Nwaubani, No. 25-9517 (4th Cir. Mar. 11, 2026) (unpublished)  |  Court: U.S. Court of Appeals for the Fourth Circuit  |  Disposition: Public admonishment under 4th Cir. R. 46(g)

1. Introduction

This attorney-discipline proceeding arises from briefing filed by attorney Eric Chibueze Nwaubani in Bolden v. Baltimore Gas and Electric Co., No. 23-2195, 2025 WL 1355304 (4th Cir. May 9, 2025), an employment-discrimination appeal. The merits panel in Bolden discovered that Nwaubani’s brief cited at least one judicial decision that did not exist and ordered a revised filing and a show-cause response. Further review revealed two additional nonexistent citations in later briefing.

The central issues before the Fourth Circuit’s Standing Panel on Attorney Discipline were:

  • whether citing three nonexistent cases in briefs submitted to the court violates the court’s discipline rule, 4th Cir. R. 46(g)(1)(c), by reference to applicable professional-conduct standards; and
  • what sanction is appropriate where the attorney denies using generative AI but the citations resemble “hallucinated” authorities and, in any event, were not caught before filing.

The parties to the discipline matter were the Respondent-attorney (Nwaubani) and the Office of Disciplinary Counsel (prosecuting counsel), with the court itself acting as the adjudicator of professional misconduct.

2. Summary of the Opinion

The Fourth Circuit imposed a public admonishment. The court held, by clear and convincing evidence, that Nwaubani violated 4th Cir. R. 46(g)(1)(c) because his conduct “with respect to this Court” violated the District of Columbia’s professional-conduct rule applicable to him—D.C. Rule of Professional Conduct 8.4(d), which prohibits “conduct that seriously interferes with the administration of justice.”

Critically, the court found it unnecessary to decide whether generative AI was used. The submission of briefs containing citations to nonexistent judicial opinions is sanctionable “no matter how it is done,” because the duty of accuracy and candor in filings remains the lawyer’s responsibility.

3. Analysis

3.1. Precedents Cited

The opinion relies on several strands of authority: (i) Fourth Circuit discipline procedure and burden of proof; (ii) due-process requirements in attorney discipline; (iii) the D.C. interpretation of “prejudicial to the administration of justice”; and (iv) emerging judicial responses to AI-related “hallucinated” citations.

A. Standard of proof and sanction framework

  • In re Liotti, 667 F.3d 419 (4th Cir. 2011): The court uses Liotti to anchor two key propositions: (1) attorney-discipline violations are proved by clear and convincing evidence; and (2) sanction selection considers aggravating/mitigating factors and “potential or actual injury,” tracking ABA Standards. This citation supplies the doctrinal “spine” for both the evidentiary threshold and the proportionality analysis.
  • In re Lattimer, 223 A.3d 437 (D.C. 2020): Cited as a corroborating comparator from the D.C. system, confirming that clear-and-convincing is standard in disciplinary matters. This matters because the substantive misconduct rule applied is from D.C., and the court reinforces that its evidentiary approach aligns with the jurisdiction whose ethical rule it is enforcing.

B. Due process in discipline proceedings

  • In re Gillespie, No. 23-1819, 2023 WL 7548181 (4th Cir. Nov. 14, 2023) and U.S. Tr. v. Delafield, 57 F.4th 414 (4th Cir. 2023): These authorities frame the constitutional minimum—notice of charges and an opportunity to be heard—also reflected in Fed. R. App. P. 46(c). The court uses them to reject Nwaubani’s claim that an error in a supplemental show-cause notice (misnaming which case was nonexistent) deprived him of fair notice. The panel’s reasoning is practical: Nwaubani identified the error, confirmed his understanding that the nonexistent citation was “Jackson,” and thus could not show meaningful lack of notice.

C. Defining “prejudicial to the administration of justice” (D.C. Rule 8.4(d))

  • In re Hopkins, 677 A.2d 55 (D.C. 1996): This is the controlling interpretive precedent for D.C. Rule 8.4(d). The court imports Hopkins’s three-part test: (1) improper conduct; (2) conduct that directly bears on the judicial process; and (3) more than de minimis taint of that process. The Fourth Circuit then maps Nwaubani’s conduct onto each element: multiple fabricated citations are improper, they are embedded in court filings (the judicial process itself), and they burden and distort adjudication beyond a trivial level.

D. AI-hallucination cases and guidance (used for context, not as elements)

The opinion surveys the accelerating body of cases addressing AI-driven errors, emphasizing that existing gatekeeping duties already cover these problems:

  • Fletcher v. Experian Info. Sols., Inc., No. 25-20086, 2026 WL 456842 (5th Cir. Feb. 18, 2026): Quoted for the proposition that “the same sanctions rules apply” to generative AI; novelty of technology does not dilute the lawyer’s responsibility.
  • Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023): Cited for the “gatekeeping role” of attorneys to ensure accuracy of filings, linking the AI problem back to familiar enforcement mechanisms (including Rule 11 in the district court context).
  • Park v. Kim, 91 F.4th 610 (2d Cir. 2024), and district-court decisions (Holmes v. The Univ. of Texas at Austin; Lexos Media IP, LLC v. Overstock.com, Inc.; Lifetime Well LLC v. IBSpot.com Inc.; Hall v. Acad. Charter Sch.): These citations function as a “trend” citation block: courts increasingly treat hallucinated citations as a serious integrity-of-process problem.
  • D.C. Bar, Ethics Opinion 388 (April 2024), Attorneys' Use of Generative Artificial Intelligence in Client Matters: Invoked to show that professional-responsibility norms already require competence, supervision, and verification when lawyers use AI tools—supporting the court’s key move of making the sanction analysis tool-neutral.
  • Kideckel v. Foreign Nation of Can., No. 1:24-cv-02907, 2026 WL 125208 (D.D.C. Jan. 16, 2026) (quoting Williams v. Cap. One Bank, N.A., No. 24-cv-2032, 2025 WL 843285 (D.D.C. Mar. 18, 2025)): Reinforces the blunt operational norm: filings may not cite authority that does not exist, “whether drafted with the assistance of artificial intelligence or not.”

3.2. Legal Reasoning

A. Jurisdictional hook: 4th Cir. R. 46(g)(1)(c)

The Fourth Circuit’s discipline power is implemented through Local Rule 46(g). Under 4th Cir. R. 46(g)(1)(c), an attorney may be disciplined for conduct before the court that violates (among other sources) “the rules of professional conduct or responsibility in effect in the state or other jurisdiction in which the attorney maintains his or her principal office.” Because Nwaubani’s principal office is in Washington, D.C., the court applies the D.C. Rules of Professional Conduct—specifically Rule 8.4(d).

B. Tool-neutral misconduct: the court declines to litigate “AI use”

Although the proceeding was triggered by concern that citations were “potentially derived from generative AI,” the court explicitly treats that as non-dispositive. The doctrinal move is important: the wrong is not “using AI,” but filing materially inaccurate citations to judicial decisions and failing to catch them through reasonable attorney review. This insulates the holding from becoming obsolete as research workflows evolve (including AI-integrated LexisNexis and Westlaw features).

C. Applying D.C. Rule 8.4(d) via In re Hopkins

Under In re Hopkins, the court finds:

  • Improper conduct: Three nonexistent cases were cited, each with citations materially different from the authorities the attorney later claimed he intended to use. The court emphasizes that a “reasonable attorney” should have discovered such errors when preparing or reviewing filings.
  • Direct bearing on the judicial process: The mis-citations appeared in briefs and were addressed at oral argument—core adjudicative materials affecting how judges evaluate claims and authorities.
  • More than de minimis taint: Nonexistent authorities waste judicial resources, distort the adversarial testing of legal propositions, and undermine the reliability of the record presented for decision. The court ties this to “advocates’ candor in submissions to the court.”

The court’s discussion of Nwaubani’s attempted “substitutions” is also significant: replacing a fabricated citation with real cases whose relevance is “tenuous at best” does not cure the original breach; it underscores that the citations were not merely typographical slips but symptomatic of a failure to ensure that authorities both exist and actually support the propositions asserted.

D. Procedural fairness: notice and opportunity to respond

Nwaubani argued a due-process violation based on an error in the supplemental show-cause notice (it mistakenly referenced Nationwide Mutual Insurance Co. v. Darden as nonexistent instead of the cited Nationwide Mutual Insurance Co. v. Jackson). Applying the notice principles reflected in In re Gillespie and U.S. Tr. v. Delafield, and the requirements of Fed. R. App. P. 46(c) and Local Rule 46(g)(4), the court rejects the claim because Nwaubani (i) recognized the error, (ii) confirmed his understanding of the true conduct at issue, and (iii) had opportunities to respond and request a hearing.

E. Sanction selection: public admonishment

Using 4th Cir. R. 46(g)(2) and the framework referenced in In re Liotti, the court chooses a public admonishment rather than suspension, monetary sanctions, or other measures. The opinion identifies limited mitigation: Nwaubani’s “self-admitted citation oversights” and “attenuated explanations” do not meaningfully mitigate. Yet the court ultimately signals proportional restraint—punishing the integrity breach while using admonishment as a clear warning to the bar that “there is no shortcut around the rules.”

3.3. Impact

  • Establishes a clear, tool-neutral discipline principle in the Fourth Circuit: Lawyers can be disciplined for citing nonexistent cases regardless of whether the source is generative AI, sloppy research, or any other workflow failure. The decision makes the verification obligation explicit in the appellate context.
  • Integrates state ethics rules into federal appellate discipline: By applying D.C. Rule 8.4(d) through 4th Cir. R. 46(g)(1)(c), the court illustrates how local attorney-discipline regimes can incorporate jurisdiction-specific ethics standards while still protecting the federal judicial process.
  • Signals heightened scrutiny of citation integrity: The opinion’s emphasis on “reasonable attorney” review and the non-trivial harm to the process suggests future panels may more readily refer similar conduct for discipline—especially where multiple hallucinated citations appear, or where “replacement” cases do not actually support the propositions advanced.
  • Practical consequence for briefing practice: The case will likely accelerate adoption of internal controls: table-of-authorities audits, spot-checking with official reporters/databases, and explicit workflows for verifying any AI-assisted research outputs before filing.

4. Complex Concepts Simplified

  • Generative AI “hallucinations”: When an AI system produces plausible-sounding but false outputs— here, judicial opinions with realistic party names and citations that do not exist.
  • “Clear and convincing evidence”: A heightened proof standard—more than “preponderance of the evidence,” less than “beyond a reasonable doubt.” It requires a firm belief or conviction in the truth of the allegations.
  • D.C. Rule 8.4(d): A professional-conduct rule prohibiting conduct that seriously interferes with the administration of justice; it is not limited to intentional fraud and can be triggered by serious failures of professional responsibility that impair court functioning.
  • 4th Cir. R. 46(g)(1)(c): The Fourth Circuit’s mechanism to discipline attorneys for misconduct in connection with the court, including violations of the ethics rules of the jurisdiction where the lawyer’s main office is located.
  • Public admonishment: A formal, public disciplinary sanction (less severe than suspension or disbarment) intended to denounce misconduct and deter repetition by the lawyer and others.
  • “De minimis” taint: A trivial or negligible impact. The court found the harm here was more than trivial because false citations consume judicial resources and distort legal analysis.

5. Conclusion

In re: Eric Chibueze Nwaubani delivers a straightforward but consequential rule for modern briefing: an attorney’s duty to ensure cited authorities exist and support the propositions asserted is non-delegable and technology-agnostic. Applying the D.C. understanding of Rule 8.4(d) through 4th Cir. R. 46(g)(1)(c), the Fourth Circuit held that filing multiple nonexistent case citations seriously interferes with the administration of justice and warrants discipline—here, a public admonishment. The broader message to appellate practitioners is clear: whether research is performed by traditional methods or AI-assisted tools, accuracy and candor in citations are foundational obligations, and failures will be addressed as threats to the integrity of the judicial process.

Note on precedential status: The court designates the decision “UNPUBLISHED” and states that unpublished opinions are not binding precedent in the Fourth Circuit. Nonetheless, the opinion is a detailed indicator of how the court is likely to treat hallucinated or nonexistent citations in future disciplinary matters.