The U.S. Court of Appeals for the Tenth Circuit has proposed a rule to address the ever-growing glut of court filings with incorrect or hallucinated case citations created by generative artificial intelligence (AI) during document creation. If implemented, it promises to make the workloads of law interns and first-year associates even heavier.
Proposed on September 18, 2026, the rule would require any court document drafted with generative AI tools to include a certification that a human reviewed the work to confirm it contains no references to misquoted or nonexistent cases in its legal analysis. As part of the rulemaking process, the proposal is open for public comment until October 18. If adopted, it would make the Tenth Circuit the first appeals court in the federal judiciary to require human verification as part of its AI disclosure requirements, with potential additional sanctions.
While tools like ChatGPT can aid in legal research, the proclivity for AI to misinterpret, misquote, and fabricate cases has unleashed a tsunami of flawed and incorrect filings in local, state, and federal courts. If the Tenth Circuit’s proposed rule goes into effect and proves effective, it may well find a home in the other 12 circuits as well.
The Ghosts You Chase, You Never Catch
Whether you consider AI a groundbreaking tool, wildly overhyped, or somewhere in the middle, the legal profession has proved eager to embrace the technology, even if it turns out to involve riding the tiger. In theory, it can feel like AI was created with the law in mind: the exacting, often tedious work of finding case histories to support a legal argument could be done by simply feeding generative AI a prompt and getting results back in a fraction of the time.
Alas, the shortcuts AI provided often turned out to be filled with thorns and, in some instances, went in unintended directions. Courts began to discover that some AI-generated court filings contained citations to cases that were misquoted, misapplied, or didn’t exist. As increasingly irked judges began disciplining attorneys, existing rules can feel somewhat behind the times in addressing the chaos created by this new technology.
Under Rule 11 of the Federal Rules of Civil Procedure, any document submitted to the court by an attorney or a pro se litigant must have evidentiary support for any factual contentions it presents. AI-generated hallucinations can look legitimate unless checked, which means they can slip into court filings. Documents submitted by self-represented litigants, which have increased significantly with the emergence of AI tools, are more likely to contain undetected mistakes. The Tenth Circuit’s proposed amendment would add an appellate-specific reminder: a person signing an AI-assisted filing would certify that a human independently verified all cited authorities.
Have You Met the New Sanctions?
The Tenth Circuit, which handles appeals from Colorado, New Mexico, Kansas, Oklahoma, Wyoming, and Utah, chose to take action after a pair of AI-related filing issues earlier this year that resulted in a fine for an attorney and a warning issued to a pro se party. The court says final 2027 rules will be posted by December 1, with the changes scheduled to take effect January 1, 2027. The rule puts the onus on the person submitting a court document, whether fully or partially created with AI tools, to independently verify the existence of all cited authorities and the accuracy of any quotes and references.
As written, the procedural rule would apply to anyone who files a document with the Tenth Circuit, including pro se applicants. In addition to existing punishments for transgressors, such as monetary sanctions, administrative discipline, and dismissal, the proposed rule would authorize the court to deem the affected issue waived or strike a noncompliant submission.
Will AI tool use by attorneys be curtailed if the rule is implemented, or would the hours needed for a human review of every AI-generated court filing become just another cost of doing business? It may become a tug-of-war between the potential to reduce legal research durations and the time spent checking for and correcting any mistakes it generates. Generative AI will likely remain part of the legal landscape. How large a role it plays and how much it’ll be trusted going forward are other matters entirely.
