Law firm escapes sanctions by having clear policy for use of AI (Leiske v. Kidd)

17 November 2026

Generative artificial intelligence (“GenAI”) tools are now widely used in everyday legal practice. GenAI tools can assist with drafting, condense and refine written material, and outline arguments in a fraction of the time it would otherwise take. At the same time, however, courts across North America have been faced with a steady stream of briefs and factums containing fabricated cases, invented quotations and mischaracterized holdings. The question is no longer whether such errors will attract judicial scrutiny but what happens when they do.

That question was central to the decision of the Delaware Court of Chancery in Leiske v. Kidd, C.A. No. 2025-0426-CDW (LWW) (November 13, 2026), in which a senior partner was sanctioned personally under Court of Chancery Rule 11 for filing a brief containing unverified GenAI output.

Law firms may wish to take note of the decision since the firm escaped sanctions for the partner’s conduct.

The underlying proceeding concerned advancement. On January 22, 2026, the plaintiffs filed an answering brief opposing the defendants’ exceptions to a Magistrate in Chancery’s final report. Five days later, the defendants’ reply brief pointed out that the brief contained fabricated quotations, inaccurate citations and hallucinated legal propositions.

The following day, the plaintiffs’ lead counsel, who had signed the brief, filed a letter acknowledging that a GenAI tool had been used and submitted a corrected brief. The letter attributed the failure to catch the errors to an oversight during a paralegal’s cite-check.

However, while the corrected brief removed quotation marks around the nonexistent quotes, it retained the inaccurate descriptions of the underlying cases.

When defence counsel wrote to the court identifying the remaining misrepresentations, the plaintiffs’ counsel responded by faulting opposing counsel for failing to “meet and confer” before alerting the court and insisting that the cited authorities were real and supported the arguments advanced.

After resolving the exceptions dispute on the merits, the court issued an Order to Show Cause concerning the GenAI issues, which directed both the signing partner and his law firm to explain why sanctions should not be imposed under the applicable court rule and the court’s inherent authority. The partner and the firm filed detailed affidavits, and the court determined the matter without an evidentiary hearing.

The court began by emphasizing that the use of GenAI in legal practice is neither forbidden nor inherently problematic, and that these tools offer substantial benefits and efficiencies. However, they must be used with the care required by the court’s Rule 11 and the Delaware Lawyers’ Rules of Professional Conduct, including ensuring the accuracy of every authority and citation presented to the court.

Under Court of Chancery Rule 11(b)(2), an attorney who presents a paper to the court certifies that, after an inquiry reasonable in the circumstances, its legal contentions are warranted by existing law. In the court’s words, “[a]n attorney’s signature on a court filing is not a procedural formality; it is a certification to the tribunal.” A fake quotation or a fabricated legal synthesis generated by an algorithm is not “existing law,” and presenting unverified GenAI output therefore violates the rule.

The court then addressed the firm’s position. Rule 11(c)(1) provides that, absent exceptional circumstances, a law firm must be held jointly responsible for violations committed by its partners, associates or employees. Drawing on Rule 5.1(a) of the Delaware Lawyers’ Rules of Professional Conduct and the American Bar Association’s Formal Opinion 512, the court asked whether the firm had made reasonable efforts to put in place measures giving reasonable assurance of compliance.

Fortunately for the law firm, the court concluded that it had. Importantly, the firm had barred the use of unapproved GenAI tools in March 2023, formed a working group in 2024 to evaluate tools and study ethical guidance, and adopted a comprehensive GenAI policy in July 2025 requiring an experienced lawyer to review GenAI output for accuracy and completeness.

The firm had also developed an accredited ethics training program that over 200 employees had completed by January 2026. As the court put it, the firm “did the right thing,” and it should not be sanctioned for an individual lawyer bypassing those controls. The firm therefore escaped sanctions.

The signing partner was not so lucky. As Chair of his firm’s AI subcommittee, he had been piloting a GenAI tool that was not approved for general firm use. Yet he used that tool to draft and condense arguments in the brief. He then delegated cite-checking to a junior lawyer and a paralegal and filed the brief without personally verifying the authorities.

That delegation was no excuse. The court held that the duty of candour and the obligation to conduct a reasonable inquiry under Rule 11 are personal and nondelegable. While practitioners routinely and appropriately rely on paralegals and junior lawyers for cite-checking, a senior lawyer who signs a brief cannot disclaim responsibility for unverified GenAI output. Where a lawyer knowingly introduces text generated by GenAI into a draft, the signatory must ensure that the underlying authorities have been reviewed and that those authorities actually support the propositions for which they are cited.

The court also rejected the argument that it was sufficient that the cases cited were real or that the overarching propositions were arguably consistent with general principles of law, noting that GenAI tools not only invent case names but also fabricate plausible-sounding quotations and stretch real decisions to stand for propositions they never addressed. If GenAI errors are detected, the court stated, counsel has an affirmative duty to review the underlying authorities and correct the errors: “It is not enough to remove quotation marks and leave a fabricated legal synthesis in place.”

As to the sanction, the court weighed the partner’s initial deflection of blame and the retained mischaracterizations against his subsequent acceptance of full responsibility, sincere contrition and long record of practice before the court. It imposed a fine of US$2,500, consistent with fines in comparable proceedings. Given the nondelegable nature of the duty, the partner was required to pay it personally. Additional mandatory continuing education was considered redundant, since the lawyer had already completed several GenAI ethics programs and led internal training.

Although this is a Delaware decision, the reasoning will resonate with litigators and their firms in any jurisdiction where courts are confronting the same problem. A lawyer who signs a factum or other document citing authorities owns every citation in it, and reliance on a cite-check by others will not shift that responsibility, particularly where the signatory knows that GenAI-generated text was used. Verification must go beyond confirming that a case exists; it requires reading the authority to confirm that any quotation is accurate and that the decision supports the proposition for which it is cited. For law firms, the decision underscores the value of a clear GenAI policy, verification requirements, restrictions on unapproved tools and documented training for members of the firm.

 

(This blog is provided for educational purposes only, and does not necessarily reflect the views of Gardiner Roberts LLP).

Author

Related Posts