In the Interest of S.M.T. and S.J.T., Children

What sanction did the court impose?

The court overruled all of the father's complaints and affirmed the trial court's agreed order. Although the appellee did not request sanctions, the court, on its own motion and citing its inherent authority to sanction improper conduct, ordered Lavespere to attend three hours of Continuing Legal Education on the dangers of the use of artificial intelligence in court filings and to certify attendance and compliance to the court within sixty days of the opinion. A footnote specifies that the three hours are in addition to the 15 hours the State Bar of Texas requires each year and must be attended, in person or remotely, rather than self-studied. The opinion's header reads "Affirmed; Nonmonetary Attorney Sanctions Ordered." The opinion imposes no monetary sanction and does not mention a bar referral. The opinion also noted the Texas Supreme Court's proposed amendments to Texas Rule of Appellate Procedure 9.1, which it said are expected to take effect October 1, 2026 and would have a signer represent that the signer "reviewed the filing and verified its accuracy, including that the legal citations, quotations, and holdings are not fabricated."

Why does In the Interest of S.M.T. and S.J.T., Children matter for law firms using AI?

The Fourteenth Court of Appeals decided the merits first, affirming an agreed custody modification order, and then turned to what it called “some discussion of appellant’s brief.” The opinion walks through five authorities page by page. Two are real cases carrying quotations the court found fabricated (In re Marriage of Harrison and In re Marriage of Swim), one is a real Texas Supreme Court decision credited with a “dual purposes” holding that “appear[s] nowhere in Lenz,” and two are real cases cited for propositions they do not address (Dickerson v. State for a request for oral argument, and In re A.S. for a contempt-enforceability rule). The court did not find that the cases themselves were invented; the defects were in what the brief said they held or said.

On the source of the errors, the opinion stops short of a finding. It states that the brief was “either (1) the result of using artificial intelligence yielding hallucination citations or (2) a deliberate attempt to mislead this court,” and adds, in the first person, “I prefer to assume the use of AI rather than intentional deception.” The opinion does not report any explanation from counsel. It frames the response around lawyers specifically: “While courts might ignore hallucination cases and quotations from pro se litigants, we cannot turn a blind eye when lawyers commit the same transgressions.”

The sanction is modest and educational: three hours of attended CLE on the dangers of AI in court filings, over and above the annual Texas requirement, with a certification due to the court within sixty days. The court imposed it although “appellee did not request sanctions,” relying on the inherent authority of Texas courts of appeals to sanction improper conduct, and pointed to the Supreme Court of Texas’s proposed amendments to Tex. R. App. P. 9.1, which the opinion says are expected to take effect October 1, 2026 and would add a signer’s representation that citations, quotations, and holdings are not fabricated. The opinion notes that the proposed rule does not change a court of appeals’ existing authority to impose sanctions.