A recent case out of the Southern District of New York (No. 24-CV-6290 (DEH) (BCM), SDNY) reads like a manual of how not to use AI. In fact, the court repeatedly found that the plaintiff’s filings bore the hallmarks of unsupervised generative-AI drafting — including hallucinated, non-existent case citations — and sanctioned the plaintiff accordingly. The AI-related findings appear across three principal decisions:

  • LiveVideo I — Report & Recommendation (the Report or the R&R), Aug. 12, 2025 (Moses, M.J.) (2025 WL 2933706 (S.D.N.Y. Aug. 12, 2025));
  • LiveVideo II — Opinion & Order adopting the R&R, Sept. 30, 2025 (Ho, J.) (2025 WL 2778110 (S.D.N.Y. Sept. 30, 2025)); and
  • LiveVideo III — Opinion & Order on sanctions, June 9, 2026 (Ho, J.) (2026 WL 1662131 (S.D.N.Y. June 9, 2026)).

1. “Telltale signs” of AI drafting (Aug. 12, 2025)

    Magistrate Judge Moses first flagged the AI problem in the Aug. 12, 2025, Report. Although the plaintiff’s Rule 59(e) papers bore counsel’s typed signature, the court observed that they were marked by odd formatting, a poor grasp of citation basics, jumbled and incomplete sentences, incomprehensible paragraphs, and random quotations from inapposite cases. The court noted that the brief included passages bearing “telltale signs of having been supplied by…an AI tool”— pointing specifically to a leftover line reading, “Here’s a revised and more coherent version of the text, with duplicate ideas removed and missing ideas added to lists.” (2025 WL 2933706 (S.D.N.Y. Aug. 12, 2025)).

    The R&R did not, however, rest its Rule 11 recommendation on the AI drafting as such. The recommended sanctions at that time were grounded in other issues, including the improper purpose behind the suit, and counsel’s abdication of his nondelegable Rule 11 duties (including allowing his ECF credentials to be used by others to file papers he did not write, review, or authorize). The AI drafting functioned as evidence of that abdication rather than as an independent basis for sanctions.

    2. Adoption of the Report (Sept. 30, 2025)

      Judge Ho adopted the Report in full, imposing the financial penalty jointly on plaintiff and its counsel, referring counsel to the District’s Grievance Committee, and entering a filing injunction restricting further motions in this case as well as future related suits in any federal district court (2025 WL 2778110 (S.D.N.Y. Sept. 30, 2025)) (The Filing Injunction). This order did not include any independent AI analysis – significantly, it made the Report’s recommendations binding and set up the contempt/sanctions posture that produced the central AI ruling.

      3. The Central AI Ruling — Fee Sanctions for AI-Fabricated Filings (June 9, 2026)

      Then on June 9, 2026, Judge Ho issued an opinion and order granting in full defendants’ motion for sanctions. In granting the motion, the court agreed with defense counsel that the action was part of a yearslong pattern of vexatious litigation and found that the challenged filings were pursued in bad faith. The court also found that, in direct violation of the Filing Injunction, the plaintiff filed a flurry of motions, and “nearly all of the motions … contain the hallmarks of artificial intelligence (‘AI’) fabrication, including non-existent hallucinated cases, formatting unchanged from that produced by AI platforms, and remnants of prompts that the drafter declined to delete before filing.” The court also noted counsel’s admission that he did not prepare, review, or file at least one of plaintiff’s motions. (2026 WL 1662131 (S.D.N.Y. June 9, 2026)). The court found the pattern continued even into the sanctions briefing itself: Plaintiff’s April 14, 2026, opposition “again appears to be drafted entirely by AI and contains hallucinated case citations.” (2026 WL 1662131 (S.D.N.Y. June 9, 2026)).

      In the operative analysis, the court held that each of the struck motions was sanctionable for three independent reasons: (1) motions filed in violation of the Filing Injunction are per se meritless; (2) the motions were repetitive of previously rejected arguments; and (3) each “appears to have been drafted by AI, contains hallucinated, non-existent quotations, and provides no real authority to support the arguments made within.” For the AI point, the court relied on the Second Circuit’s decision in Park v. Kim,[1] which referred an attorney to the Grievance Committee for a Rule 11 violation for preparing filings with an AI tool and failing to ensure their accuracy. (2026 WL 1662131 (S.D.N.Y. June 9, 2026)).

      The June 9, 2026, sanctions were imposed under 28 U.S.C. § 1927 and the court’s inherent power, not Rule 11. Under both, the court may award fees on clear evidence that the offending party’s claims were entirely without color and were brought in bad faith — and bad faith may be inferred from conduct so meritless it must have been undertaken for an improper purpose such as delay or harassment. AI-driven, hallucination-laden, authority-free character of the filings supported both the “without color” and “bad faith” findings.

      Having determined that sanctions were warranted, the court applied the Second Circuit’s lodestar framework and awarded $85,056 in attorneys’ fees and costs, payable within 21 days, and again referred plaintiff’s counsel to the Grievance Committee.

      What This Means for Attorneys Using Generative AI

      There are several notable takeaways from this series of decisions. First, AI misuse is corroborating evidence of sanctionable conduct, and not alone sanctionable activity. The court never treated the sloppy “use of AI” as itself sanctionable. The sanctionable conduct was filing frivolous, injunction-violating papers containing fabricated authority. The AI findings mattered because they confirmed that the filings were untethered from any genuine legal research and that counsel had not discharged his personal duty to verify what was filed under his name. This tracks the emerging judicial consensus reflected in Park v. Kim: the violation lies in failing to check AI output, not in improperly using the tool.

      Second, the attorney’s nondelegable certification duty is the linchpin. The recurring theme across all three decisions is that counsel signed, or allowed to be filed under his credentials, papers he did not write, review, or authorize. Hallucinated citations are the visible symptom; the Rule 11 problem is the absence of the “inquiry reasonable under the circumstances” that an attorney’s signature certifies. An attorney may not insulate himself by attributing AI-generated work to someone else.

      Finally, the AI-fabrication conduct persisted despite a $10,000 penalty, a Grievance Committee referral, and the Filing Injunction — and even appeared in the very brief opposing the fee motion. That unbroken pattern drove the court from a modest deterrent penalty to a full fee-shifting award and a second disciplinary referral. The case illustrates how repeated, uncorrected AI misuse compounds the exposure for both party and counsel. The decisions underscore that filing AI-assisted papers without verifying every citation and quotation may expose counsel to Rule 11 sanctions, § 1927 and inherent-power fee awards, and disciplinary referral — and that courts now recognize specific “hallmarks” of unsupervised AI drafting as red flags.

      [1] Park v. Kim, 91 F.4th 610 (2d Cir. 2024).

      Print:
      Email this postTweet this postLike this postShare this post on LinkedIn
      Photo of Kathryn C. Cole Kathryn C. Cole

      Kathryn C. Cole is a shareholder and commercial litigator who represents businesses, financial institutions, and individuals in complex disputes across federal and state courts, arbitration, mediation, and proceedings before federal agencies and regulatory bodies. Her practice spans contract claims, product liability, tort litigation…

      Kathryn C. Cole is a shareholder and commercial litigator who represents businesses, financial institutions, and individuals in complex disputes across federal and state courts, arbitration, mediation, and proceedings before federal agencies and regulatory bodies. Her practice spans contract claims, product liability, tort litigation, consumer class actions, and securities class actions, serving clients ranging from major pharmaceutical companies to closely held businesses. Katy brings a combination of courtroom advocacy, appellate experience in the U.S. Court of Appeals for the Second Circuit, and deep knowledge of eDiscovery strategy that allows her to manage high-stakes litigation from preservation through trial and appeal.

      Katy is a Certified E-Discovery Specialist (CEDS) and a recognized authority on electronic discovery and data retention. She guides clients through every phase of the eDiscovery process, including identification, preservation, collection, review, and production of electronically stored information. She works closely with clients to develop defensible discovery plans and retention policies, implement technology assisted review platforms, and leverage analytics to reduce costs and risk. She is a founding member of and prolific contributor to Greenberg Traurig’s eDiscovery Watch blog, where she regularly addresses emerging issues such as generative AI in litigation, text message preservation, and spoliation.

      Before entering private practice, Katy clerked for the Hon. Richard C. Wesley at both the New York Court of Appeals and the U.S. Court of Appeals for the Second Circuit. She has argued multiple appeals in the Second Circuit and the Second Department Appellate Division, and she has in-house experience through a secondment with a major technology company. Her understanding of both the courtroom and the corporate legal department gives her a practical perspective that informs her litigation strategy and client counseling.

      Katy is actively involved in shaping civil practice in New York. She serves on the Office of Court Administration’s CPLR Advisory Committee and the New York State Advisory Committee on Civil Practice. She is a past president of the Eastern District of New York Chapter of the Federal Bar Association and a Fellow of The New York Bar Foundation. She has been listed in The Best Lawyers in America for Commercial Litigation and she received the Long Island Business News “Top 50 Women in Business” award in 2024.

      Read more about Kathryn C. ColeKaty's Linkedin Profile
      Show more Show less
      Related Posts
      AI hallucinations 1200x450 ppt
      From Hallucination to Sanctions: The Potential Costs of Unverified AI Citations
      October 20, 2026
      Chat Tools for Business Communication-shutterstock_2519088537
      Mastering the Chat Matrix: Technical E-Discovery Considerations for Collaborative Communications Platforms
      June 16, 2026
      Binary code background vector illustration, blockchain and legal, made of triangle shape, crushing and fading in the dark-shutterstock_1299146458
      Court Sanctions Highlight Potential Risks of Using Unchecked AI in Litigation
      March 10, 2026