A federal judge in Michigan criticized the Trump administration’s Department of Justice for using a nonexistent court decision in an immigration detention case. Court records indicate the citation may have been created by artificial intelligence.
Chief U.S. District Judge Hala Y. Jarbou of the Western District of Michigan highlighted this issue on July 16. The case involved an ICE detainee who contested the government’s automatic stay of a bond order. The judge, appointed by Trump, eventually dismissed the case as moot. The stay had expired, and the government revealed Izzeddin Ahmed Abdulgahffar Daghra could be released with a $35,000 bond.
Questionable Citation in DOJ Filing
Judge Jarbou’s order examined a citation in a DOJ filing. Government lawyers claimed the Sixth Circuit ruled that federal courts can’t reassess evidence behind a bond decision or challenge the Immigration Judge’s discretion, referencing a nonexistent case, Taylor v. Hott. This dispute arose amidst ongoing friction between the judiciary and the Trump administration. Federal judges have been scrutinizing and reprimanding government actions more frequently.
The judge stated that attorneys using AI tools must meticulously verify their work before submitting documents to the court. She described AI-generated legal research as contributing to a wave of cases with fabricated authorities, where briefs contain “fake, hallucinated cases” that don’t exist.
“It should be obvious that any attorney who uses AI must scrupulously review its work product to ensure that the cited cases exist and that the citations accurately and fairly represent the underlying case law,” Judge Jarbou wrote.
While she didn’t impose sanctions, Judge Jarbou emphasized that future government filings must avoid using nonexistent legal authorities.
Previous Instances of AI-Generated Legal Filings
The Michigan case is not isolated. In April, U.S. Magistrate Judge Robert T. Numbers II from North Carolina reprimanded former federal prosecutor Rudy Renfer. Renfer admitted to utilizing generative AI in a court brief, which contained false quotations, incorrect citations, and mischaracterizations of case law.