By now the story of a lawyer caught citing fake AI-generated cases is almost a genre. A brief gets filed, the opposing side or the judge tries to pull a cited case, and the case is simply not there. It never existed. A chatbot invented it, complete with a plausible name, a plausible citation, and a plausible holding, and a busy attorney pasted it into a federal filing without checking. What makes the June 8, 2026 ruling out of the Northern District of Mississippi different is the symmetry of it. This was not one careless lawyer getting caught. It was both sides of the same case, independently, doing the exact same thing, and getting caught together.
The case is Withers v. City of Aberdeen, a contract and attorney-fee dispute brought by Tom Withers III against the City of Aberdeen, Mississippi. U.S. District Judge Sharion Aycock, presiding over the matter, found that filings from both the plaintiff's team and the city's team contained citations to nonexistent cases. The result was a set of sanctions that hit four attorneys at once, barred two of them from practicing in the district for two years, disqualified the entire roster, and blew up the trial schedule. It is one of the cleanest illustrations yet of how a single bad habit, trusting a machine that confidently makes things up, can take down everyone in the room.
What The Judge Found On Both Sides Of The Table
The thing judges keep running into with AI-hallucinated citations is that they are designed to pass a glance. A fabricated case looks like a real one. It has the format, the reporter volume, the page number, the party names that sound like the kind of parties who would litigate that issue. You only learn it is fake when you try to read it and there is nothing to read. In Withers, multiple filings on both sides carried exactly this kind of phantom authority, and Judge Aycock did not treat it as a clerical slip. She treated it as sanctionable conduct under Federal Rule of Civil Procedure 11, the rule that requires an attorney signing a filing to certify that its legal contentions are warranted.
What gives the ruling its weight is that Aycock refused to let anyone hide behind ignorance of the technology. One explanation offered to the court, that a lawyer had been unaware AI could hallucinate at all, she rejected outright, calling it insufficient and incredulous. A practitioner in 2026 does not get to be surprised that a chatbot invents cases, not after years of headlines and prior sanctions across the federal courts. The duty to confirm that a cited case is real is not a new duty created by AI. It is the oldest duty in legal practice, and the tool did not erase it.
The Penalties, Named And Counted
The sanctions were specific, and they were not symbolic. Kathryn Young Williams, one of the out-of-state attorneys admitted pro hac vice for the city, drew the largest fine at $3,500. Kathleen Wilson, the out-of-state pro hac vice attorney for Withers, was fined $2,500. Shauncey Hunter Ridgeway, also representing Withers, and Mark McClinton, representing the city, were each fined $1,000. The two attorneys who had been admitted specially to appear in this single case, Williams and Wilson, were additionally barred from appearing before the Northern District of Mississippi for two years.
Beyond the money and the bar, all four attorneys were disqualified from the case entirely. The trial that had been scheduled was canceled, and the proceedings were paused so the litigants could secure new counsel. For a client, that is the part that stings outside the courtroom: a contract dispute that was ready for trial got reset to near zero, with the people who were supposed to be advocates removed and the calendar wiped, all because the filings on every side leaned on cases that a language model dreamed up. The court also referred the matter to the relevant state bar associations, meaning the professional consequences may not end with the fines.
The Duty That Cannot Be Outsourced
The most quotable, and most important, line in the ruling is also the most basic principle in the profession. Aycock wrote that a lawyer's duty to verify the law they cite is absolute, and that it cannot be outsourced to technology or delegated to co-counsel. That is the whole case in one sentence. The problem was never that these lawyers used software. The problem was that they let the software stand in for the one thing they are personally, non-transferably responsible for: confirming that what they tell a court is true.
She put a finer point on the mechanism of the failure, too. Their practice of blindly relying on technology, she wrote, resulted in the hallucinatory citations contained in their respective filings. Blindly is the operative word. Not using AI. Using it blindly, treating fluent output as verified output, skipping the step where a human opens the cited case and reads it. It is the same failure pattern this site documents in every other corner of the AI economy, where confident, well-formatted output gets trusted in proportion to how convincing it sounds rather than how often it is correct.
And the court made clear this was not its first encounter with the problem. Aycock noted that the bench was yet again burdened with addressing AI hallucinations in court filings. That phrase, yet again, is the quiet alarm in the whole document. Individual judges are now absorbing a recurring tax on their time, re-checking citations that should never have needed checking, because attorneys keep shipping machine-generated work without the human verification step that is supposed to be the entire point of hiring a lawyer.
Why This One Matters More Than The Usual Sanction
Most AI citation scandals are one-sided, which lets everyone tell a comforting story: that careless lawyer got caught, the rest of us are fine. Withers refuses that comfort. When both the plaintiff's counsel and the defense counsel, working independently, in the same case, commit the same error, the issue stops looking like a few bad apples and starts looking like a structural change in how legal work gets done. The tool is now woven into drafting on every side of the aisle, and the verification discipline that is supposed to backstop it has not kept pace.
This is the same hallucination problem this site has tracked everywhere else AI gets deployed faster than its output can be checked. We have documented how AI-hallucinated court citations have been flooding courts and drawing increasingly brutal sanctions, and how the same fabrication dynamic shows up far from any courtroom when unverified AI output quietly kills enterprise projects in budget review. The Mississippi ruling is the legal system's version of the same lesson, delivered with the bluntest instrument a court has: fines, a multi-year bar, disqualification, and a canceled trial.
The defense, as always, will be that the attorneys did not mean to deceive anyone, that they were busy, that the tool is new and the norms are still forming. Aycock's answer is the right one, and it is unsentimental. The duty to verify is absolute. It predates the technology and survives it. A lawyer who signs a filing is vouching for it, and a chatbot's confident invention is not a defense, it is the thing the lawyer was supposed to catch.
The Verdict
Four attorneys, on both sides of Withers v. City of Aberdeen, filed federal briefs citing cases that a chatbot invented and they never bothered to read. The cost was $3,500 in fines at the top end, a two-year bar from the district for the two out-of-state lawyers, disqualification for all four, a canceled trial, and a referral to the state bars. The duty to verify, the judge said, is absolute and cannot be outsourced to technology. AI did not break that rule. Lawyers trusting AI blindly did.
Has an AI tool produced fabricated facts, citations, or sources that caused real consequences for you? Tell us what happened.