Federal judges scold Severna Park law firm for AI-generated fake citations
Two partners at Quinn Patton filed at least five briefs containing fabricated case law quotes generated by an artificial intelligence editing tool, prompting federal judges in Maryland and Washington, D.C., to demand explanations and threaten sanctions.
Managing partners Donald Quinn and Katherine Patton, both admitted to the bar in 2021, used a closed AI system to edit legal documents. The tool produced false citations and misquoted case law - what the legal profession calls "hallucinations." Three of the problematic filings landed before U.S. District Judge Stephanie A. Gallagher, who issued separate orders this March requiring the attorneys to explain the errors.
Quinn said the firm discontinued the AI tool once it became aware of the software's editing issues. Patton declined to comment, citing ongoing litigation in her civil cases.
Pattern of violations across two jurisdictions
This wasn't Quinn's first encounter with judicial displeasure over AI hallucinations. Last October, U.S. District Judge Rudolph Contreras in Washington, D.C., issued a warning after Quinn filed a brief containing fabricated citations. Contreras wrote that "further violations will not be tolerated and may result in sanctions and/or referral to a disciplinary committee."
Less than a month later, the violations continued. Patton filed a motion in a Maryland discrimination case with inaccurate quotations. Quinn filed another brief in a D.C. district case that opposing counsel flagged as relying on "case citations that do not exist." Two more filings followed in sexual harassment and retaliation cases.
Judge Gallagher made clear the firm's problems extended beyond individual briefs. She wrote that her concerns "extend to counsel's entire firm" and rejected a partially corrected brief from Patton, stating that "attorneys are not entitled to a second bite at the apple" after filing documents with hallucinated citations.
Judges grapple with verification burden
Most federal courts do not outright ban generative AI in legal filings. But judges increasingly face the time-consuming task of verifying citations that attorneys should have caught before submission.
James Rubinowitz, a New York attorney who lectures on AI in courts at Cardozo School of Law, said the problem will worsen without real consequences. "Unless the courts start leveling some real consequences against attorneys, there's no incentive for the attorneys to slow down," he said.
Rubinowitz pointed to a recent case in which Sullivan & Cromwell, a major Wall Street firm, apologized to a federal judge after filing a brief with inaccurate citations. He said five problematic filings from Quinn Patton aren't surprising - attorneys believe AI tools save time, then lack the hours needed to verify what the software produced.
"It's no excuse for lawyers who do this because it's such a black eye on anybody who practices law," Rubinowitz said.
The core problem with language models
Large language models were not designed to verify facts. They generate plausible-sounding text, not truthful responses. A citation that looks correct to a skimming attorney may be entirely fabricated.
Rubinowitz said citation errors are costing judges significant time and resources. Courts can no longer assume citations are accurate and must verify attorneys' work independently. Clients ultimately bear the cost through delayed cases and attorney fees spent correcting preventable errors.
When attorneys discover their filings contained fake citations, the ethical path is clear: notify the client, consult an ethics attorney, and disclose the violation. "Fall on your own sword," Rubinowitz said.
For legal professionals using AI tools, understanding these risks is essential. Learn more about AI for Legal work and consider the AI Learning Path for Paralegals, which covers document preparation and citation verification in an AI-assisted environment.
Your membership also unlocks:
