Senior U.S. District Judge Sharion Aycock had seen enough. In a breach-of-contract dispute over unpaid legal fees between the city of Aberdeen, Miss., and Louisiana attorney Tom Withers III, she discovered something extraordinary. Lawyers for both parties had submitted court filings laced with citations to cases that never existed. All generated by artificial intelligence.
The revelation came after the court ordered both sides to produce copies of the authorities they cited. None could. One side referenced a nonexistent 1971 Mississippi Supreme Court decision. The other leaned on three phantom federal rulings. The trial? Scrapped. The lawyers? Removed from the case. Two barred from appearing in the Northern District of Mississippi for two years. Fines totaling $8,000.
“This case presents the Court with an unusual scenario—attorneys for both litigants engaged in similar sanctionable conduct.”
Judge Aycock wrote those words in her 23-page sanctions order issued June 8, 2026. The Northern District of Mississippi, she added, was “yet again ‘burdened with addressing AI hallucinations in court filings.'” The phrase carries weight. This wasn’t her court’s first encounter with the problem. It won’t be the last.
Kathleen M. Wilson represented Withers. She admitted using generative AI to draft her response. Kathryn Y. Williams appeared for the city. She used an AI tool for legal research. Both claimed ignorance of the large language models’ tendency to invent precedents. The judge didn’t buy it. One lawyer, she noted, had relied on the technology without oversight for at least six months. Previous warnings in another matter had gone unheeded.
“This court finds that explanation to be insufficient and incredulous,” Aycock declared. Short. Direct. Devastating. The sanctions hit all four attorneys involved, with the two primary AI users facing the harshest penalties. Disqualification. Suspension from the district. Financial hits scaled by degree of fault, from $1,000 to $3,500 each.
But the real sting lies elsewhere. Two clients paid for representation. Instead, their lawyers essentially let chatbots argue against each other. AI on one side. AI on the other. Neither side caught the fabrications. The court did. Lawyer Rob Freund spotted the mess early and called it a “comedy of AI errors.” He wasn’t wrong. Clients funded artificial intelligence to litigate against itself.
The 404 Media report first brought the order to wider attention. It laid out the timeline. The admissions. The judge’s frustration. Days later, major outlets picked it up. The New York Times detailed how the federal judge punished all four lawyers, canceled the civil trial and removed them from the matter. Business Insider framed it as judges losing patience with AI mistakes flooding courtrooms. The Mississippi Free Press offered local context on the Aberdeen dispute.
This incident doesn’t stand alone. Far from it. A database maintained by researcher Damien Charlotin tracks dozens of AI hallucination cases in courts worldwide. Recent U.S. examples include a June 12, 2026, Texas matter where a pro se litigant submitted three fabricated cases, and a Georgia schools dispute days earlier that led to a vacated order. Pennsylvania’s Commonwealth Court saw at least 13 filings with confirmed or suspected AI errors in 2025 alone. Judges there have begun interrupting arguments to demand whether briefs contain “AI hallucinations.”
Earlier high-profile cases set the pattern. The 2023 Mata v. Avianca saga in New York, where a lawyer cited six nonexistent cases generated by ChatGPT, resulted in $5,000 sanctions and mandatory education. That ruling warned the profession. Many ignored it. By 2025, documented incidents had surged sevenfold according to trackers. Mike Lindell’s lawyers in Colorado paid $3,000 each in 2025 for a filing with more than two dozen errors, including fake citations. A California appeals court hit one attorney with $10,000 for 21 fabricated quotes out of 23 in a brief.
And judges aren’t immune. Staff in at least two federal chambers have produced rulings later withdrawn after AI hallucinations surfaced in the text. The problem spreads. It wastes time. It erodes trust. It forces courts to act as fact-checkers for the very professionals sworn to uphold accuracy.
Courts Respond With Stronger Measures
Standing orders now require disclosure of AI use in many districts. Local rules multiply. Bar associations issue opinions. The American Bar Association’s Formal Opinion 512 from 2024 offers guidance on generative tools. Yet enforcement varies. Some judges impose fines. Others disqualify counsel mid-case, as seen in an Alabama federal matter involving a large firm where monetary sanctions proved “ineffective” at deterrence.
In the Mississippi case, Aycock went further. She didn’t just fine. She cleared both legal teams entirely. New counsel must start over. The original trial date vanishes. Costs mount for the parties. All because lawyers treated AI output as reliable without verification. One continued using the tool even after the judge highlighted fake cases in an earlier filing. That detail particularly galled the court.
Legal technology has advanced. So have the risks. Large language models excel at pattern matching. They falter on precision. They invent. They hallucinate. Lawyers who paste first and verify never pay the price alone. Their clients do. The justice system does. Opposing counsel wastes hours chasing ghosts.
But here’s the shift. Courts no longer accept ignorance as an excuse. “In an era of rampant unverified AI usage within the legal field, this case presents a prime example of the risk associated with serving as a rubber-stamp,” Aycock wrote. The message lands hard. Verify everything. Read your filings. Know your sources. Or face disqualification, suspension, fines and public embarrassment.
Recent coverage shows the trend accelerating. Spotlight PA reported on Pennsylvania judges questioning attorneys directly about AI in briefs. NPR tracked rising penalties across the system, noting even state supreme courts have grilled lawyers over fabricated citations. The pattern holds. Early scandals failed to slow adoption. Sanctions now stack higher.
Technology won’t disappear from legal practice. Its proper role requires human oversight at every step. Research with AI? Fine. Draft language? Possible. Submit without reading every word and checking every citation? Career risk. The Mississippi episode drives that home with unusual clarity. Both sides failed. The court noticed. Everyone lost.
Future cases will test these boundaries further. Some lawyers will adapt, building verification protocols and disclosure habits. Others won’t. Judges like Sharion Aycock stand ready. Their tolerance for AI slop in the record has reached its limit. The profession must catch up. Accuracy still matters. Human judgment remains irreplaceable. Anything less invites the hammer.
WebProNews is an iEntry Publication