Court considers sanctions after AI-assisted brief includes factitious citations

A court ordered Rich to show cause in Clark v. Social Security, Commissioner of after a brief prepared with generative artificial intelligence included factitious citations, according to CourtListener RECAP dockets: generative AI and sanctions or hallucinations. The order says the court was considering sanctions, including a fine. The docket entry was filed on September 24. It also says Rich used AI in preparing the plaintiff’s brief and responded to the show-cause order.

The significance is not that AI was used. It is that the filing still had to survive ordinary legal scrutiny. The court’s response treats unreliable citations as a problem attached to the filed work and the lawyer responsible for it, not as a problem erased by the software that helped produce it. That is the operating requirement for AI-native legal work: faster or cheaper production has value only if the finished work remains accurate enough to carry legal consequences. AI does not eliminate the consequences of bad legal work.

Clark is a concrete instance of that rule, not proof that every AI-assisted filing fails or that every court will impose the same penalty. The order establishes that a court connected AI-assisted preparation with factitious citations strongly enough to require an explanation and to consider sanctions. It does not establish that AI caused every defect in the brief, that the lawyer did not review any of it, or that a particular sanction will follow. Those stronger conclusions would require a final ruling addressing the citations, the review process, and the sanction itself. The distinction matters for AI-native firms: a company may describe its system as capable of producing finished legal work, but the relevant test is whether the work withstands review after it is filed. A product description is not a substitute for that result. The same boundary appears in the surrounding court materials. CourtListener: AI and sanctions (opinions) describes Douglas v. Deutsche Bank National Trust Co. as a cautionary tale about misuse of AI and its consequences for attorneys. Its entry for Cervantes v. Bianco says the court disapproved of improper use of AI tools that apparently contributed to a defect. Those entries support the narrower point that AI use does not displace professional or procedural consequences; they do not turn those individual matters into a general finding about all AI legal systems.

The next fact to watch is the ruling on the show-cause order in Clark v. Social Security, Commissioner of. A sanction, a fine, or an order declining to impose one would answer a narrower question than the headline claim, but it would show how the court evaluates the explanation and the review of the AI-assisted citations. A separate filing offers a useful comparison: in Macias v. CCADC, CourtListener RECAP dockets: generative AI and sanctions or hallucinations reports that OpenAI Codex assisted with the document and that the signer certified personal checking of the AI-assisted portions, including citations. Whether that certification is accepted as meaningful review, and whether Clark produces a sanction, will show whether disclosure and human checking are treated as controls or merely as statements made after defective work reaches the court. We will score the model on that observable result, not on the promise that AI makes legal work finished.