If AI produces factitious citations, courts can impose sanctions

On September 24, in Clark v. Social Security, Commissioner of, the court ordered Rich to show cause why she should not be sanctioned, including with a fine, after a plaintiff’s brief prepared with generative artificial intelligence included factitious citations, according to a CourtListener RECAP docket. The filing says Rich used AI in preparing the brief and that the court’s order followed the inclusion of those citations. The immediate event is not an AI policy statement or a warning from a vendor. It is a court proceeding that turns defective legal work into a possible sanction.

That is the narrow but important point for AI-native legal work: automation does not remove the consequences attached to a bad filing. The work still reaches a court, still carries citations that can be checked, and still exposes the responsible lawyer to a process that may include a fine. The Clark filing does not establish that AI caused every defect in the brief, or that the court has decided the final sanction. It does establish that AI use did not prevent the court from responding to the resulting citations as a professional and procedural problem.

Other court materials place the Clark matter in the same direction without making it larger than it is. In Douglas v. Deutsche Bank National Trust Co., a published order calls misuse of AI a cautionary tale and links it to consequences for attorneys, according to CourtListener: AI and sanctions (opinions). In Cervantes v. Bianco, an opinion says the court disapproves of improper use of AI tools that apparently contributed to a problem, according to CourtListener: AI and privilege (opinions). Those materials support the proposition that AI-assisted legal work remains subject to judicial scrutiny. They do not prove that every AI-assisted filing is defective, that every defective filing merits a fine, or that an AI-native firm has failed.

The stronger reading would require more than a show-cause order and references to AI use. It would require a final sanctions order identifying the conduct that violated the applicable court’s standards, the sanction imposed, and the connection between the AI-assisted process and the defective citations. It would also require facts about who reviewed the brief, what review occurred, and whether the same defect would have appeared without AI. None of that is established here. The order sanctions no one yet; it begins a process for deciding whether sanctions are warranted. A related firm, product, or delivery model is not named, so the event says nothing about the performance of AI-native law firms as a category.

The next fact to watch is the court’s disposition of the Clark show-cause process: a sanctions order, a fine, or an order declining sanctions. The useful details will be the court’s stated basis, whether the citations were verified before filing, and whether responsibility turns on AI use, inadequate review, or both. A final order imposing a sanction would strengthen the practical claim that AI-assisted legal work carries ordinary legal consequences when it fails. A final order declining sanctions would narrow the lesson to judicial scrutiny and away from punishment. We should score the model on that order, not on the presence of AI in the filing alone.