A Broader Admonition to the Bar

On May 5, 2026, the U.S. District Court for the District of Maine sanctioned plaintiff's counsel in Fuller v. Hyde School for filing two briefs that contained AI-generated citations the attorney had not independently verified. The order is significant beyond its facts: the court paired its individual sanction with what it expressly called "a broader admonition to the Bar regarding the need for caution, verification, and transparency whenever generative AI is used in litigation." (Fuller v. Hyde Sch., No. 2:25-cv-00354-SDN, slip op. at 5 (D. Me. May 5, 2026).)

The conduct

Counsel filed an opposition to a motion to dismiss in November 2025. The brief cited cases for unsupported propositions, contained quoted language that could not be located in the cited sources, and included inaccurate pincites. (Id. at 1–2.) Counsel then filed a "Notice of Errata" attempting to correct the citations — but, the court found, the Notice "introduced additional inaccuracies in the challenged citations" and "repeated and extended some of the same problems instead of fully remedying them." (Id. at 2, 4.)

Counsel did not disclose her use of AI — identified only as "either Claude or ChatGPT" — until April 2026, after the court issued an order to show cause. (Id. at 2.) She acknowledged at that point that she had not conducted a line-by-line citation check before filing. (Id.)

What the court called out

The court treated the delayed disclosure of AI use as a discrete aggravating factor, separate from the underlying verification failure. Counsel "did not candidly disclose her use of AI in her initial filings and did so only after the Court issued a show cause order, further frustrating the Court's ability to understand the source of the errors and implicating her duty of candor." (Id. at 5.) That framing — non-disclosure as a candor problem owed to the tribunal, alongside the Rule 11 verification failure itself — is the bar-relevant development in this order.

The court imposed a public admonition, mandatory CLE on generative AI in legal practice, required firm-level procedures to prevent recurrence, and a certification of compliance within 45 days. (Id. at 6.) It also struck the original opposition from the docket and ordered an amended response. (Id.)

What this asks of bar associations

Fuller joins Mata v. Avianca and Park v. Kim — both expressly cited by the court — as part of a documented body of federal sanctions orders in which attorneys submitted AI-generated content without verification. (See id. at 3.) What Fuller adds is a court's explicit treatment of non-disclosure during the proceeding as part of the misconduct, not merely as a procedural lapse but as a candor obligation distinct from the underlying citation errors.

Bar guidance on AI to date — including ABA Formal Opinion 512 and several state-level advisories — has emphasized competence, supervision, and confidentiality. Few formal opinions yet address what contemporaneous disclosure of AI assistance requires when an attorney's AI-assisted citations are challenged in active litigation. The court in Fuller did not articulate a rule, but it documented the gap: counsel had no record of verification to produce when the errors were challenged, and no disclosure of AI use to anchor the court's inquiry. The absence of both is what extended the proceedings and triggered the show-cause order.

A bar guidance framework that addresses contemporaneous documentation — what an attorney should have on file before signing an AI-assisted brief, and what to disclose when challenged — would give practitioners a clear standard against which to measure their own practices, and would support the verification-and-transparency duty the Fuller court has now admonished the Bar to observe.