If AI helps write a brief, a court can still demand an explanation

On September 24, 2026, in Clark v. Social Security, Commissioner of, the court ordered Rich to show cause why she should not be sanctioned, including with the imposition of a fine, for including factitious citations in a plaintiff’s brief. CourtListener RECAP dockets: generative AI and sanctions or hallucinations reports that the brief used generative artificial intelligence and that Rich’s response acknowledged AI use in its preparation. The order puts the consequence next to the workflow: AI was used to prepare the filing, and the filing became the subject of a sanctions inquiry.

AI does not eliminate the consequences of bad legal work. The Clark order is a direct example because the court did not treat the presence of AI as an answer to the problem in the brief. It treated the citations as a reason to demand an account from the lawyer associated with the filing. For an AI-native legal model, that is the relevant standard: delivering a brief faster or with less human labor does not change what happens when the work contains citations a court calls factitious. The client receives legal work that must survive scrutiny, not a process exemption because software helped produce it. A separate CourtListener: AI and sanctions (opinions) entry for Douglas v. Deutsche Bank National Trust Co. describes misuse of artificial intelligence as a cautionary tale about its consequences for attorneys, reinforcing the same point without establishing a general rule.

The Clark order establishes a show-cause proceeding tied to factitious citations in an AI-assisted brief. It does not establish that AI caused every defective citation, that Rich was sanctioned, or that a fine was imposed. It also does not establish that every lawyer using AI faces the same result. The stronger reading would require a final order identifying the defective citations, explaining the connection between the AI-assisted preparation and the filing, and imposing or declining a sanction. That distinction matters. A court’s demand for an explanation is already a consequence, but it is not the same thing as a final penalty or a ruling that makes AI use itself wrongful. The position is therefore exemplified, not proved as a universal proposition.

The next observable fact is the court’s final disposition of the Clark show-cause proceeding: whether it imposes a fine or another sanction, identifies the citations at issue, or ends the matter without a penalty. That order would distinguish scrutiny from punishment and show whether the consequence turned on AI use, on the factitious citations, or on both. A useful comparison is the filing in Macias v. CCADC, where the document says OpenAI Codex assisted in its preparation and includes a personal-review certification covering the AI-assisted portions and all citations. CourtListener RECAP dockets: generative AI and sanctions or hallucinations reports that certification; it does not report the court’s later assessment of whether the control worked. We will treat a final Clark order as the fact that determines how far this example reaches.