A federal appeals court is moving to tighten its rules on artificial intelligence after judges encountered court filings containing fabricated cases, false quotations and other AI-generated legal errors.
The U.S. Court of Appeals for the Tenth Circuit proposed a new rule September 18 that would require lawyers and self-represented litigants to certify that a human reviewed filings prepared with generative AI before they are submitted to the court.
Under the proposal, anyone using generative AI to help prepare a brief, motion or other filing would have to independently verify the authorities cited, check quotations and citations, review the legal analysis and confirm compliance with court rules and professional obligations.
The Denver-based court hears federal appeals from Colorado, Kansas, New Mexico, Oklahoma, Utah and Wyoming. If adopted, the change would put a specific human-review requirement into the circuit’s rules as federal courts continue confronting AI-generated mistakes in legal filings.
The proposal follows two recent Tenth Circuit cases involving fabricated or inaccurate legal authorities.
In February, the court sanctioned Maryland attorney Kusmin L. Amarsingh after finding seven nonexistent cases in a brief filed in an appeal involving Frontier Airlines. Judges also found two real decisions that did not support the quotations or legal arguments attributed to them.
Amarsingh acknowledged using ChatGPT while researching and drafting the appeal. The court ordered her to pay $1,000 toward Frontier’s legal costs and referred the decision to Maryland attorney disciplinary authorities.
Another case reached the court in June. Self-represented litigant Chapter Kris Jackson disclosed using AI-based tools to help prepare filings that judges found contained quotations and legal propositions missing from the cited cases. The panel warned that future use of hallucinated case law could result in sanctions.
The proposed rule focuses on the accuracy of what reaches the court. Lawyers already carry responsibility for the documents they sign and submit, including the factual and legal support behind their arguments. The new language would add an express certification that AI-assisted work received independent human review.
Generative AI can produce polished legal writing while inventing authorities, altering quotations or attaching the wrong legal proposition to a real case. A filing may appear complete until a judge, clerk or opposing lawyer checks the underlying source.
Those errors can create consequences beyond an embarrassing correction. Courts can impose sanctions, strike filings, refer lawyers for professional discipline and require payment of costs caused by defective submissions.
The Tenth Circuit proposal would expressly allow judges to strike a filing, treat an affected issue as waived, impose monetary sanctions, order payment of expenses and attorney fees, begin disciplinary proceedings or, in some circumstances, dismiss or affirm an appeal.
AI is also becoming increasingly embedded in legal research platforms, drafting software and document systems used by attorneys. The proposed rule would leave lawyers free to use those tools while making the person submitting the filing responsible for verifying the final product.
Other federal appeals courts have considered similar restrictions. The Fifth Circuit proposed an AI certification requirement in 2023 that would have required lawyers to disclose whether generative AI was used and confirm that AI-generated material had been checked.
After receiving comments arguing that existing professional and procedural rules already made lawyers responsible for the accuracy of their filings, the Fifth Circuit dropped the proposal in 2024.
The proposed Tenth Circuit rule centers instead on the finished filing. Lawyers would have to certify that a human verified the authorities, quotations, citations, and legal analysis submitted to the court, regardless of which AI tool helped prepare the document.
Public comments on the proposed rules are due October 18. The court plans to publish its final 2027 rules by November 30, with approved changes scheduled to take effect January 1, 2027.
A California judge has ordered a lawyer for State Farm to pay $999.99 after finding seven nonexistent case citations in pretrial filings, errors the attorney said stemmed from her use of an artificial intelligence legal research tool.Los Angeles Superior Court Judge Elizabeth L. Bradley imposed the sanction on Jacquelene Robinson,... Read More »
OpenAI is calling for new U.S. and international AI safety standards after disclosing six incidents in which its artificial intelligence models concealed mistakes, used credentials without authorization, uploaded information to the public internet, and communicated outside approved channels.The ChatGPT maker released the cases September 16 under a new framework for... Read More »
OpenAI, Anthropic, Google and SpaceXAI have been hit with a federal antitrust lawsuit accusing the rival artificial intelligence companies of illegally agreeing to slow development of ChatGPT, Claude, Gemini and Grok in the name of AI safety.Four paying subscribers filed the proposed class action September 18 in the U.S. District... Read More »
A federal judge’s AI-assisted order that named nonexistent people, cited material outside the court record, and misstated Mississippi law is now at the center of a Fifth Circuit fight over whether he should continue presiding over the case.Mississippi is asking the U.S. Court of Appeals for the Fifth Circuit to... Read More »
A New Mexico defense lawyer has been fined $5,000 after using ChatGPT to help draft a murder appeal, failing to check what it produced and filing a brief that included four witnesses who never existed, false testimony, and misrepresented legal authority.The New Mexico Supreme Court held Santa Fe attorney Stephen... Read More »
