Last month, the US District Court for the Northern District of Mississippi handed down a scathing judgment admonishing and sanctioning attorneys for both the plaintiff and defendant in the case of Tom Withers III v City of Aberdeen. The plaintiff was not suing the local authority in Scotland, but a city located in Monroe County in the US state of Mississippi. The action related to unpaid fees by the defendant to an attorney who, for better or for worse, did not represent himself but employed another attorney.
AI Hallucinations
In the course of the proceedings, the City of Aberdeen sought for the action to be dismissed without inquiry into the facts and lodged an appropriate application. Mr Withers lodged opposition and both sets of attorneys (two for the plaintiff and two for the defendant) lodged written submissions. All four lawyers signed the relevant documents lodged on behalf of the parties. The documents contained a number of citations of cases which did not exist and the four attorneys admitted this resulted from their use of AI without properly checking its output. A separate document for the plaintiff also included a quote from a non-existent case.
The Sanctions Hearing
The Court summoned all four attorneys to a hearing to determine what sanctions should be imposed on them for this misconduct. Two of the attorneys (the lead ones for the parties) admitted responsibility for drafting the documents. One attorney had used AI for research purposes. The other had used generative AI to draft her submissions to the Court. Neither had verified the output. The other two attorneys were not informed of their colleagues' use of AI but signed the submissions without checking the authorities.
Because of the issue and the hearing on sanctions, the trial hearing in the action set down for 23 March 2026 had to be abandoned and postponed to a different date (yet to be fixed).
The lead attorney for the plaintiff admitted that her single-practitioner law firm had no AI policy in place. She had been using an AI drafting tool for legal documents for about six months prior to the case in question. She admitted she had used it for the purposes of legal documents submitted in other cases. She claimed that she did not know that the tool could hallucinate cases, which the Court found "incredulous". It was of serious concern to the Court that the attorney continued to misuse the AI tool after she was put on notice about the violations in the present case. Two months after the sanctions hearing, she was sanctioned by another court in another matter in respect of two other submissions which included hallucinated cases and statutes. The Court found that this was a failure on the attorney's part to end that practice despite giving the Court an assurance that she would.
The lead attorney for the defendant, on the other hand, practised as a partner in a Texas law firm which had an AI policy in place. The policy mandated that all AI output had to be verified by the attorney. The Court found it "particularly egregious" that the attorney failed to do so. Additionally, the continuing professional development training which the attorney had attended throughout 2025 included topics on AI use.
The Sanctions Imposed
The Court reviewed the applicable standards and case law and determined that the attorneys had abused the judicial process. US case law specifically held that courts "have the inherent power to impose sanctions for abuse of the judicial process" and that "[s]ubmitting a brief riddled with fabricated [case law] is such an abuse". In respect of the use of AI, the Court referred to previous case law noting that: "AI is a powerful tool, that when used prudently, provides immense benefits…[h]owever, while "[g]enerative technology can produce words," it cannot attach "... sincerity, truth, or responsibility to what it writes. That remains the sacred duty of the lawyer who signs the page." The Court was also clear that even if the legal propositions advanced were accurate, this does not diminish the misconduct of bringing hallucinated cases to the Court's attention. It was no basis to reduce any appropriate sanction. The Court referred to another US case noting: "any sanctions discount on this basis would amplify the siren call of unverified AI for lawyers who are already confident in their legal conclusion. This court will have no part of that."
The Court found that the two attorneys who had carried out the drafting of the submissions had a higher degree of culpability as they had acted in bad faith. The Court was of the view that sanctions as to the ability of these attorneys to practise before the Court were appropriate. The two lead attorneys' ability to practise before the District Court for the Northern District of Mississippi was therefore barred for a period of two years. Additionally, the lead attorney for the plaintiff was fined $2,500 and the lead attorney for the defendant, $3,500. The other two attorneys were disqualified from appearing in that case and were fined $1,000 each. The parties were given time to procure new representation.
Conclusion
This case is a stark reminder of the vigilance which every solicitor ought to employ when using AI tools in the practice of law. As has been noted many times before, the use of AI does not abrogate the responsibility to deploy critical thought. In a sense, AI ought to be treated as a person who provides a view of the law - it cannot be taken at its word and must be rigorously verified.
We saw this verification being required in a case a few weeks ago, where Tamires Camal Taquidir instructed a debt recovery law firm powered by AI to recover unpaid compensation in the amount of £7,000. Albeit through the use of a human barrister, it was noted that the AI technology employed was capable of preparing papers for trial (including witness statements). It is unclear what human input was provided in relation to the preparation of these documents, but one would expect that it was of critical importance. Ms Taquidir was able to recover the debt and successfully fend off a counterclaim. However, it is difficult to imagine that this result would have been possible without critical thinking on the part of humans being deployed both in respect of the preparation of the case and the required advocacy. Only when the two elements, AI technology and human skill and judgment, come together can meaningful advantage be taken of the opportunities this tool presents.
Whilst these examples demonstrate both the potential and the pitfalls of AI in legal practice, we have been fortunate in Scotland not to have reported cases involving legal practitioners misusing AI technology in their dealings before the Court. However, this is no reason to be complacent - the more the technology advances, the more comfort some practitioners might feel in using AI output without independently verifying it. As Sheriff MacRitchie observed recently in Your Home Partners v Kellichan & Hood [2026] SC KDY 34 (where no solicitors were involved), these cases demonstrate "the dangers of a reckless reliance on Artificial Intelligence by any party without verifying that the same is genuine, and the potential for a party being found in contempt of court in circumstances such as these, even if done in good faith."
This article was co-authored by Josh Chambers
