
A federal judge in Mississippi removed both legal teams from a civil case after discovering each side filed AI-hallucinated citations. The sanctions order in Withers v. Aberdeen is the first known instance of a court catching both parties relying on fake AI-generated material in the same case.
A federal judge in Mississippi kicked both legal teams off a civil case after discovering that attorneys on each side had submitted court filings containing fake citations generated by artificial intelligence. The June 8 sanctions order in Tom Withers III v. City of Aberdeen is the first known instance of a court catching both parties relying on AI hallucinations in the same case, according to the ruling.
The case is stayed to let the plaintiff find new counsel. If the plaintiff cannot, the case will be dismissed.
Judge Roy Percy of the U.S. District Court for the Northern District of Mississippi ordered the removal of all attorneys from the case after reviewing explanations from both sides. The court also imposed additional sanctions, though the order did not detail them.
AI hallucinations occur when large language models like ChatGPT or Gemini generate fabricated legal citations, quotations, or case summaries that look real but do not exist. The problem has become a persistent headache for courts since late 2023, when the first high-profile case – a lawyer submitting fake cases generated by ChatGPT – drew national attention.
Since then, dozens of similar incidents have surfaced. Most resulted in mild rebukes or modest fines. Judges have generally accepted arguments that attorneys were unfamiliar with the technology. That leniency is fading.
In this case, the court found that both sides had submitted filings containing AI-hallucinated material. Neither side caught the errors on their own. Neither side flagged the other's mistakes. The court discovered the problem during its own review.
The ruling did not specify how many fake citations each side filed or which parts of the record were tainted. The court said it weighed the nature and materiality of the hallucinations before deciding on consequences.
Legal ethics experts said the decision to remove both legal teams – rather than issue fines or refer them to bar disciplinary boards – signals a shift in how courts view AI-related misconduct.
"This is a shot across the bow," said Michael Downey, a legal ethics partner at Armstrong Teasdale in St. Louis. "Judges are running out of patience with the 'the computer did it' defense. Removing counsel from a case is a serious step that directly affects the client."
The American Bar Association issued Formal Opinion 512 in July 2024, stating that lawyers have an ethical duty to supervise AI tools and verify their output. The opinion made clear that ignorance of how generative AI works is not a defense.
"Two years after that opinion, claiming you didn't know AI could hallucinate is no longer credible," Downey said.
The case also raises a strategic dilemma for attorneys who discover AI hallucinations in their own work or in an opponent's filing. Reporting the other side's errors could expose their own. Staying quiet risks the court finding out on its own, which could lead to harsher penalties for concealing the problem.
"If you detect AI hallucinations and say nothing, you are violating your duty of candor to the court," said Renee Knake Jefferson, a legal ethics professor at the University of Houston Law Center. "The calculus changes when both sides are in the same boat. The duty does not disappear."
The court did not address whether either side knew about the hallucinations before filing. The sanctions order focused on the fact that the material was submitted and not caught.
Legal observers expect more cases involving AI hallucinations on both sides as attorneys continue to use generative AI without adequate oversight. The Mississippi case provides a template for how courts might handle the situation going forward.
"This is not going to be the last time a judge sees this," Jefferson said. "The question is whether other courts follow this approach or go further."
Prepared with AlphaScala research tooling and grounded in primary market data: live prices, fundamentals, SEC filings, hedge-fund holdings, and insider activity. Each story is checked against AlphaScala publishing rules before release. Educational coverage, not personalized advice.