US Court Imposes Sanctions on Lawyers for Submitting AI-Generated Fabricated Case Law in Patent Dispute

Introduction

In a significant ruling on professional ethics in the age of artificial intelligence, the United States District Court for the District of Kansas has sanctioned five attorneys for violating Federal Rule of Civil Procedure 11 by filing court documents containing fabricated and misrepresented legal citations generated by unverified AI tools. The case, Lexos Media IP, LLC v. Overstock.com, Inc. , Case No. 22-2324-JAR, involves a patent infringement lawsuit where plaintiff Lexos accused defendant Overstock of infringing patents related to cursor image modification on websites. U.S. District Judge Julie A. Robinson, presiding over the matter, issued a detailed memorandum and order on February 2, 2026, publicly admonishing the attorneys and imposing monetary fines totaling $12,000, while revoking one attorney's pro hac vice admission. This decision underscores the court's stance that while generative AI is not inherently problematic in legal practice, the failure to verify its output constitutes a serious breach of professional standards, as echoed in contemporaneous litigation news reports. The sanctions highlight growing judicial concerns over AI "hallucinations" in court filings, serving as a cautionary tale for lawyers navigating technology in high-stakes patent litigation.

Case Background

The underlying dispute originated in 2022 when Lexos Media IP, LLC, a plaintiff known for enforcing similar patents across multiple jurisdictions, filed suit against Overstock.com, Inc., alleging infringement of three patents concerning the dynamic modification of cursor images on websites to enhance user interaction. The case progressed through standard pretrial stages, including a claim construction hearing on September 21, 2023, where the court defined key terms like "cursor image" to include not just a movable display element but one indicating where user input could be received.

By mid-2025, the litigation intensified with Overstock filing a motion for summary judgment on June 13, 2025, alongside two Daubert motions to exclude expert testimony—one targeting Lexos's technical expert, Samuel Russ, Ph.D., for failing to apply the full claim construction of "cursor image," and another against damages expert Justin Blok. Lexos's response, due on July 7, 2025, became the flashpoint. Attorneys from Buether Joe & Counselors, LLC (Texas-based, handling the case on contingency), SethLaw PLLC (co-counsel), and local counsel Fisher, Patterson, Sayler & Smith, LLP (Kansas) collaborated on the filings. Notably, lead attorney Eric W. Buether had passed away in October 2024, leaving Christopher M. Joe to assume leadership.

The legal questions at the heart of the sanctions were not the merits of the patent claims but the attorneys' compliance with Rule 11, which requires a reasonable inquiry into the law before filing. The defective filings—Lexos's opposition to the Russ exclusion motion (Doc. 194) and parts of the summary judgment response (Doc. 193)—included numerous nonexistent cases, fabricated quotations, and misstated legal principles, all traced to unverified use of ChatGPT by co-counsel Sandeep Seth. Overstock's reply brief exposed the errors on July 15, 2025, prompting Lexos to seek leave to correct the documents on July 23, 2025. Judge Robinson denied substantive corrections to avoid prejudice but allowed excision of the fakes, striking Doc. 194 initially. On December 15, 2025, she issued an order to show cause why the attorneys should not face sanctions, leading to the February 2026 ruling after reviewing their declarations.

This timeline reflects a broader pattern in Lexos's patent enforcement strategy, with similar suits against entities like eBay and Amazon, often involving the same counsel team. The sanctions episode arose amid escalating expert disputes, where accurate legal authority was crucial to defending the admissibility of Russ's testimony on infringement.

Arguments Presented

The sanctions ruling did not revisit the underlying patent arguments but focused on the attorneys' show-cause responses defending their conduct under Rule 11. Sandeep Seth, who admitted sole responsibility for using generative AI, argued in his declarations that personal hardships—caring for his terminally ill mother and aunt during July 2025—distracted him, leading to an "isolated poor decision" as a "novice" in AI tools. He claimed ChatGPT was used as a "shortcut" for 10th Circuit and Federal Circuit case law supporting non-exclusion of experts for minor claim construction errors, insisting he intended to verify outputs but failed due to emotional strain. Seth emphasized that 95% of the brief was human-drafted, downplayed intent to deceive, and noted no billing to the client for the flawed sections. He requested no sanctions for co-counsel, taking full blame, and pledged never to use AI for case law again.

The Buether Joe attorneys—Christopher M. Joe (lead and managing partner), Kenneth P. Kula (senior counsel), and associate Michael W. Doell—collectively denied knowledge of Seth's AI use, citing their firm's strict no-AI policy for court filings. Joe attested he never reviewed the documents before signing, having assumed leadership post-Buether's death but delegating to Seth and Doell. Kula, on family vacation abroad during filing, worked only on the unrelated Blok Daubert response and understood Seth had overall responsibility. Doell admitted drafting non-defective sections of the summary judgment response and making minor edits to Seth's work but, as a junior associate, was not tasked with substantive legal review, especially given Seth's 30+ years of patent experience. They argued their reliance on Seth and internal policies absolved them, with no bad faith or awareness of errors.

Local counsel David R. Cooper acknowledged reading the briefs on filing day but not cite-checking, having previously caught a citation error in an earlier submission. He relied on Seth and Doell for accuracy, expressing remorse and detailing his firm's new AI prohibition policy, including verification requirements and a self-imposed 12-month ban on sponsoring pro hac vice admissions. Cooper highlighted his 25 years of practice in the district and compliance with local rules requiring meaningful participation.

Opposing perspectives emerged implicitly through the court's lens: Overstock's motions emphasized Russ's flawed analysis as fatal to infringement claims, while the defective responses undermined Lexos's defense by citing phantom authority like nonexistent Hockett v. City of Topeka and misquoted Federal Circuit cases such as Flexuspine, Inc. v. Globus Medical, Inc. . The attorneys' defenses centered on lack of intent and delegation, but the court rejected these as insufficient under Rule 11's objective reasonableness standard.

Legal Analysis

Judge Robinson's 36-page order meticulously dissects the Rule 11 violations, applying an objective standard of reasonableness to determine if a competent attorney would have filed the documents. Rule 11(b)(2) mandates that legal contentions be warranted by existing law or nonfrivolous extensions thereof, following a reasonable pre-filing inquiry. The court found unanimous violations among the five signatories for submitting briefs riddled with "hallucinated" authority: eleven instances in Doc. 194 alone, including fabricated quotes from real cases like i4i Ltd. P'ship v. Microsoft Corp. (misstating jury credibility over exclusion for claim construction errors) and entirely invented decisions like Hockett v. City of Topeka (actually a Social Security appeal under a different name).

The analysis draws on precedents affirming Rule 11's application to AI misuse. Citing Mata v. Avianca, Inc. (S.D.N.Y. 2023), the court noted that fake opinions are not "existing law," rendering arguments frivolous. Tenth Circuit cases like Predator Int’l, Inc. v. Gamo Outdoor USA, Inc. (2015) reinforced the nondelegable duty to verify, rejecting "blind reliance" on co-counsel ( Wadsworth v. Walmart Inc. , D. Wyo. 2025). The ruling distinguishes AI's permissibility from verification failures, aligning with the other source material: "the use of generative AI in the legal profession is not inherently problematic, the failure to verify its output constitutes a serious breach of professional standards."

Key principles applied include deterrence as Rule 11's core purpose ( Cooter & Gell v. Hartmarx Corp. , 1990) and proportionality in sanctions ( White v. Gen. Motors Corp. , 10th Cir. 1990). The court weighed relative fault: Seth's unawareness of AI risks despite publicized cases (over 500 documented by 2025) and failure to seek extensions or share burdens aggravated his role. Buether Joe partners Joe and Kula's zero review breached supervisory duties under ABA Model Rules 5.1 and 5.3, while Doell's subordinate position mitigated hers. Cooper's partial review as local counsel invoked D. Kan. Rule 83.5.4's "meaningful participation" requirement, emphasizing local counsel's vouching role.

The decision navigates related concepts like subjective bad faith (unnecessary under objective Rule 11) versus objective inquiry, and AI policy versus ethical competence. It references ABA Formal Op. 512 (2024) and state bars (Texas Op. 705, California guidance), urging training on AI limitations. Broader context includes surging AI sanction cases ( Johnson v. Dunn , N.D. Ala. 2025), with courts adopting standing orders (e.g., E.D. Tex., W.D.N.C.). This ruling clarifies that in patent suits, where precise claim construction and expert reliability are pivotal, unverified AI erodes judicial efficiency and opponent resources.

Key Observations

Judge Robinson's order is replete with incisive excerpts emphasizing the gravity of the violations and the unchanging demands of legal ethics amid technological evolution. Key quotations include:

• On the core violation: "There is no question that citing to a nonexistent case, attributing a nonexistent quotation to an existing case, and misstating the law violates Rule 11(b). Such reliance does not constitute 'existing law' and certainly does not provide to the Court a nonfrivolous argument for extending, modifying, or reversing existing law or for establishing new law."

• Highlighting AI's perils: "The violation here is due to the failure to verify that the cases generated by ChatGPT actually exist and confirm that they stand for the propositions for which they are cited... As attorneys transition to the world of AI, the duty to check their sources and make a reasonable inquiry into existing law remains unchanged."

• Addressing delegation: "Blind reliance on another attorney can be an improper delegation of this duty and a violation of Rule 11... When an attorney gives another permission to sign on their behalf without reviewing the document, it is not only actionable under Rule 11 but also a possible violation of a state’s ethical rule of competence."

• On deterrence and context: "A reasonably competent attorney filing documents in court should be aware of the pronounced, well-publicized risks of using unverified generative AI for legal research and the ethical obligations associated with signing a court filing without checking it for accuracy."

• Final admonition: "The attorneys' collective failure to devise a system of checking their filings for accuracy in the face of the well-publicized need to do so requires sanctions in order to deter repetition of this conduct by them and by other attorneys who are similarly situated."

These observations, drawn verbatim from the judgment, encapsulate the court's reasoning, blending empathy for personal circumstances with firm enforcement of professional norms. They integrate seamlessly with external reports, such as the litigation news snippet affirming that AI use itself is unproblematic but unverified outputs breach standards.

Court's Decision

The court unequivocally held that all five attorneys violated Rule 11(b)(2), issuing public admonitions and tailored sanctions to deter recurrence while considering fault levels. Sandeep Seth received the harshest: a $5,000 fine, revocation of pro hac vice admission (given ample remaining counsel for Lexos), mandatory self-reporting to Texas and California disciplinary authorities with a copy of the order, and a deadline of February 28, 2026, to file a certificate detailing firm procedures for verifying legal authority—addressing his novice status and policy gaps.

Kenneth P. Kula and Christopher M. Joe each faced $3,000 fines for abdicating review duties; Joe additionally must implement and certify enhanced firm policies by February 28, 2026, including AI training and verification under ABA guidelines, reflecting his managerial role. David R. Cooper incurred a $1,000 fine, acknowledging his review but failure to verify, bolstered by his firm's proactive policy updates and voluntary pro hac vice hiatus. Michael W. Doell escaped further penalties beyond admonishment, with the court's sympathy for his untenable junior position and the firm's remedial measures sufficing.

The order also struck specific erroneous passages from Doc. 193 (pages 38-39 on claim exclusivity) and reinforced prior striking of Doc. 194. Fines are payable to the court's registry within 14 days. Practically, this weakens Lexos's opposition to excluding Russ's testimony, potentially paving Overstock's path to summary judgment on noninfringement without technical expertise. Implications ripple beyond this case: it signals heightened scrutiny in patent litigation, where AI could streamline but not shortcut claim analysis or Daubert challenges. For future cases, attorneys must prioritize human verification, fostering AI competence training and inter-firm protocols—especially in co-counsel setups. This decision, amid a national surge in AI-related sanctions (e.g., fines up to $31,000 in Lacey v. State Farm , C.D. Cal. 2025), may spur bar associations and courts to standardize AI guidelines, protecting judicial integrity while embracing innovation. Ultimately, it reaffirms that technology serves the law, not supplants the lawyer's duty.

(Word count: 1,478)