Attorneys on both sides of a contract dispute paid a steep professional price after a Mississippi federal judge removed all four of them from an active case following the discovery of AI-hallucinated citations in their court filings.
- A June 2026 sanctions order in the Northern District of Mississippi removed all four attorneys from a case after AI-generated, fabricated citations were found in filings by both legal teams
- The two drafting attorneys received two-year bans from the court, monetary fines, and bar referrals; local counsel on both sides were disqualified and fined for failing to review the documents they signed
- The ruling reinforces that Rule 11 obligations cannot be outsourced to technology or delegated to co-counsel, and that ignorance of AI hallucination risks is no longer a valid defense
An unusual scenario with a clear warning
In June 2026, Senior United States District Judge Sharion Aycock of the Northern District of Mississippi issued a sanctions order in Withers v. City of Aberdeen that is hard to ignore: every attorney of record on both sides of the case was removed from the litigation on the same day. The case itself was a straightforward breach-of-contract dispute over legal fees. What made it anything but ordinary was the conduct that preceded the sanctions.
In November 2025, the court reviewed briefs from both parties and could not locate several of the legal authorities cited. The citations were not merely misquoted or miscited; they did not exist. The court identified six fabricated case citations across three filings and, on December 10, 2025, issued an order to show cause directing all counsel of record to explain their conduct.
At a January 20, 2026, hearing, the attorneys acknowledged the source of the problem. Kathryn Y. Williams, lead counsel for the City of Aberdeen, acknowledged using an AI research tool for legal research without independently verifying the output. Kathleen M. Wilson, lead counsel for plaintiff Tom Withers III, admitted using an AI drafting tool called “First Drafts” to prepare her filing and failing to verify the citations it produced. Both local counsel, Shauncey Hunter Ridgeway and Mark C. McClinton, admitted signing filings they had not reviewed (Bloomberg Law, Lawyers Sanctioned Over AI Use on Both Sides in Federal Case, June 9, 2026).
A profession rapidly adopting AI, with uneven safeguards
The Withers sanctions did not occur in a vacuum. AI adoption across the legal profession has accelerated sharply in recent years. According to the 2025 Clio Legal Trends Report, 79% of legal professionals were using AI tools in their practice (Clio, 2025 Legal Trends Report). At the same time, the ABA’s 2024 Legal Technology Survey found that concerns about AI hallucinations remain the top barrier to adoption, with approximately 75% of attorneys citing accuracy concerns as a reason for hesitation (American Bar Association, 2024 Artificial Intelligence TechReport). The gap between adoption and caution is precisely where professional liability lies.
That same report found that more than half of legal professionals said their firm had no AI policy in place or that they were unaware of one (Clio, 2025 Legal Trends Report). The attorneys at Withers illustrated what that gap looks like in practice.
What the court found, and why it mattered
Judge Aycock’s analysis drew a clear line between the two drafting attorneys and the two local counsel, though she sanctioned all four under Rule 11 of the Federal Rules of Civil Procedure. Rule 11 requires that any attorney who signs a filing certify, based on a reasonable inquiry, that the legal contentions are warranted by existing law. The court found that submitting fabricated citations is an absolute violation of that duty, irrespective of intent.
For Williams and Wilson, the court went further, invoking its inherent authority to sanction conduct amounting to bad faith. Williams knew her firm’s AI research tool was not designed to produce Mississippi case law, and her firm had a written policy requiring independent verification of AI output, yet she disregarded both. Wilson had used the “First Drafts” tool for about six months without verification, and the court found violations across multiple filings, establishing a pattern of conduct. The court cited the recent Fifth Circuit decision Fletcher v. Experian Information Solutions, holding that submitting briefs containing fabricated authority constitutes an abuse of the judicial process and that pleading ignorance of AI risks is, in the court’s words, no longer a viable excuse.
The conduct of both local counsel illustrated a separate but equally serious risk: the danger of serving as a “rubber stamp.” Ridgeway and McClinton each signed filings without reviewing the legal citations they contained. The court found that this practice violated not only Rule 11 but also the Northern District of Mississippi’s Local Rule 83.1, which holds resident attorneys responsible for the conduct of proceedings brought by out-of-state co-counsel admitted pro hac vice.
The Withers case was not the only proceeding in which Wilson’s misuse of AI had consequences. In April 2026, a federal bankruptcy court in the Western District of Louisiana separately sanctioned Wilson for submitting two additional filings containing AI hallucinations in a matter filed in March 2026, two months after she had apologized to Judge Aycock’s court and promised to change her conduct.
Sanctions that left no one standing
The resulting sanctions were sweeping. Williams and Wilson each had their pro hac vice admissions revoked and were barred from appearing before the Northern District of Mississippi for two years. Williams was ordered to pay a $3,500 fine; Wilson was ordered to pay $2,500 and to attend a continuing legal education course on AI with an ethics component, as she was the only attorney among the four who had not previously attended such a CLE. Ridgeway and McClinton were disqualified from further participation in the case and each ordered to pay a $1,000 fine. The court also directed the clerk to transmit copies of the sanctions order to the Mississippi Bar, the Louisiana Bar, and the Texas Bar, as well as to the judges presiding over related proceedings involving Wilson’s subsequent conduct.
Both parties were given 60 days to retain new counsel, with a warning that the plaintiff’s case would be dismissed if he failed to do so.
What litigation teams should take from this ruling
The Withers order reinforces several obligations that apply to every attorney who uses AI tools in connection with court filings, regardless of firm size or jurisdiction.
First, verifying every citation against a primary source is not optional. Westlaw, LexisNexis, or the official records of the relevant court must confirm that a cited case exists and that it says what the filing claims it says. Relying on an AI tool’s output, even a legal-specific one, does not satisfy this duty. As the court observed, the Rule 11 obligation “cannot be outsourced to technology or delegated to co-counsel.”
Second, firm AI policies are only as effective as their enforcement. Williams’ firm had a written policy requiring independent verification, yet she violated it. The court found that violation particularly aggravating. Firms that invest in AI governance documentation should also ensure that compliance is a routine, supervised part of the filing process.
Third, local counsel and supervising attorneys bear independent liability for the documents they sign. Ridgeway and McClinton did not draft the tainted filings, were unaware of their co-counsel’s use of AI, and did not personally use AI. None of that shielded them from sanctions. Signing a document is a certification under Rule 11, and that certification requires a reasonable inquiry, including a review of the cited legal authorities.
Fourth, the consequences of AI hallucinations extend beyond the individual filing. In Withers, the court could not resolve a pending summary judgment motion. The scheduled trial was canceled, and the litigation ground to a halt for months while both parties scrambled to find replacement counsel. The reputational and practical costs to clients in such circumstances are substantial and largely invisible to the attorneys who caused them.
The Withers ruling is part of a growing body of decisions signaling that courts have moved beyond the warning stage. Attorneys who have not yet built a disciplined verification workflow into their AI use are not merely taking a professional risk. As this case shows, they are also risking their clients’ interests.
You might also like:
GenAI Prompting for Legal Professionals
The Modern Attachment Revolution: How Cloud Links Are Reshaping Electronic Discovery
The Evolution of Cloud Forensics: Challenges and Solutions in Cloud-Based Investigations