Chakma v. Sushi Katsuei, Inc.
U.S. District Court, Southern District of New York · S.D.N.Y. · New York bar guidance
Conduct
Defense counsel used a LexisNexis AI tool to help draft letters to the court, did not check the output, and waited over six months to seek to replace the false submissions.
Consequence
Counsel pays $1,710 of class counsel's fees plus a $1,000 penalty; the clients separately pay $47,761.88 for discovery and class-notice misconduct.
Lesson
Here the court found unverified AI use to be bad faith despite sympathetic personal circumstances, but kept the penalty low for genuine remorse and an apparently isolated incident.
Verified September 30, 2026
- Citation
- Chakma v. Sushi Katsuei, Inc., No. 23 Civ. 7804 (KPF), Opinion and Order (S.D.N.Y. May 19, 2026) (Failla, J.), ECF No. 142
- Decided
- May 19, 2026
Summary
In a wage-and-hour class action against two Sushi Katsuei restaurants and two individual defendants, class counsel moved for sanctions based on three categories of post-certification conduct. The third concerned legal authority in defense counsel's letters to the court. Judge Katherine Polk Failla's order describes a July 8, 2025 letter that cited Rule 23 and OConner v. Agilant Solutions for a hearsay point, and OConner and Haider v. Lyft on agency, although neither case addresses those issues, and that cited "§ 3.01 Certification Is Crucial Step In Class Actions" with no indication of its source. At a September 16, 2025 hearing, defense counsel acknowledged that she had used LexisNexis's AI tool to help construct multiple submissions and then failed to review the final versions, and recognized that certain citations, including the "§ 3.01" reference, resulted from AI hallucinations. The court remarked at the hearing that in letters filed July 1, 8, and 11, 2025, "all of the cases that are cited seem to be irrelevant" to defendants' points. Defendants did not seek to replace the submissions until January 21, 2026, after the sanctions motion raised them.
- AI tool:
- LexisNexis AI tool (defense counsel "acknowledged that she had used LexisNexis's AI tool to help construct multiple submissions to the Court")
- Sanction amount:
- Against defense counsel for the AI-generated false submissions: $1,710 of class counsel's fees plus a $1,000 penalty payable to the Clerk
What sanction did the court impose?
The court granted the sanctions motion in large part. On the AI-related submissions, it found that defense counsel's "use of generative AI without the necessary checks nonetheless constitutes bad faith," while expressing sympathy for serious personal circumstances she described, and ordered her to pay class counsel $1,710 in fees for litigating the false submissions. It also imposed a $1,000 penalty on defense counsel, payable to the Clerk within 30 days, which it described as "on the low end" of what courts have imposed because of her personal challenges and remorse the court found genuine, and because, quoting Benjamin v. Costco, "it appears that this is the only time" she used an AI platform to replace her own legal judgment. Separately, the court ordered defendants to pay $46,980 in fees and $781.88 in costs for delayed class-list and payroll discovery and for interference with the class-notice process. Defense counsel may now move to replace the AI-generated submissions. The court noted that defense counsel had already sent class counsel checks for $1,900 in fees attributable to her and for the $781.88 in costs.
Why does Chakma v. Sushi Katsuei, Inc. matter for law firms using AI?
Most of this opinion concerns the defendants’ discovery conduct after class certification, but a distinct part addresses defense counsel’s use of AI. The court’s account comes largely from the September 16, 2025 hearing transcript: counsel acknowledged using LexisNexis’s AI tool to help construct several submissions, admitted failing to review the final versions, and recognized that some citations were hallucinations. The court was skeptical that failing to review was the only problem, observing that in the July 8 letter “there’s no case that [Defendants] cit[ed] that actually stands for the proposition” for which it was cited.
The order sets out Rule 37, the court’s inherent power, and 28 U.S.C. § 1927 as the governing authorities, and for this category it found bad faith. It quoted Braica v. Frankowski that “courts in this circuit have repeatedly found that presenting AI-generated hallucinations as valid caselaw constitutes subjective bad faith,” and adopted Braica’s language that counsel “made no attempts to check” whether her AI-generated letters were accurate and used “these misleading hallucinations to advance h[er] case.” It then shifted class counsel’s fees for litigating the false submissions to her personally.
The $1,000 penalty reflects deliberate calibration. Citing Benjamin v. Costco’s survey of $1,500 to $15,000 sanctions, the court set a lower figure because of her personal challenges, genuine remorse, and the apparent absence of prior AI misuse, while stating that those circumstances “do not excuse” the conduct. Firms documenting compliance may wish to consider that, in this case, a delay of more than six months in correcting admitted false authority was one of the facts the court emphasized.
Implications for your firm
Operational steps a firm reading this case may wish to consider documenting. Strategic and rule-application calls belong to your firm's attorneys.
- Verify output from AI features inside established research platforms. Counsel's hallucinated citations here came from a LexisNexis AI tool.
- Document prompt correction of known errors. The court stressed that counsel admitted the false authority in September 2025 but did not move to replace the submissions until January 2026, after the sanctions motion.
- Review letters and short submissions as carefully as briefs. The hallucinated authority here appeared in letters to the court about discovery and class notice.
- Consider that the court treated sympathetic personal circumstances as mitigating the amount of the penalty, not as excusing the conduct.