Artificial intelligence (AI) has moved rapidly from novelty to near necessity across the legal sector as firms fight to keep up with the ever-changing technological landscape. However, recent cases continue to demonstrate how the hurried, and unchecked, use of AI can derail proceedings, mislead courts and jeopardise the standing of legal professionals.
The leading authority on the use of AI is R. (on the application of Ayinde) v Haringey LBC [2025] EWJC 1383 (Admin). This case sets out comprehensive guidance and emphasised the professional and ethical duties of legal practitioners when using AI for legal research.
The court made clear that generative AI tools are not capable of conducting reliable legal research. They can generate entirely fabricated case law and citations. It is imperative that legal professionals using AI verify any output against authoritative sources before using it in court or advising clients.
Despite the guidance in the above judgment in June 2025, there have been a number of instances since where legal professionals have fallen short of these obligations.
Ndaryoyimvire v Birmingham City University [2025] 10 WLUK 719
The claimant made an application to amend their particulars of claim on 10 July 2025. The application, signed in the name of the claimant’s solicitor Mr Newton, referred to two authorities which the defendant’s solicitors requested copies of as they could not find them. It transpired that these authorities did not exist and the claimant’s solicitor asked the court to disregard the first application, submitting an amended second application on 14 July 2025. The second application was struck out and the defendant’s solicitor sought wasted costs. When questioned on the inclusion of the false authorities, Mr Newton explained that the first application was still in draft when it was submitted and had been submitted in error without his knowledge. The first application had not been signed by Mr Newton personally but had been signed by another member of staff. It was also stated that the claimant’s firm had relied upon a piece of AI-assisted research software, which had provided the citations.
On the basis that it was largely a failure of management at the firm, rather than a failure of Mr Newton as an individual solicitor, it was decided that consideration of matters such as contempt proceedings or a referral to the police would be disproportionate.
However, in the judgment in October 2025, HHJ Charman took the opportunity to reiterate that lawyers who cite fictious cases face serious consequences and judges should take a tough line given the growing prevalence of the problem. A wasted costs order was made against the solicitor’s firm.
Choksi v IPS Law LLP [2025] EWHC 2804 (Ch)
In this case the claimant submitted an application for summary judgment. In response the defendant firm of solicitors served two witness statements from its managing partner, the second of which appeared to have been drafted, at least partially, using AI.
The claimant complained that the defendant’s evidence was unsatisfactory and deficient for a number of reasons. The statement cited a number of authorities (which is, of course, not the purpose of a witness statement) and, in any event, several of which had wrong citations or names, and did not support the proposition for which they were cited. In some cases, the authorities were completely irrelevant or did not exist.
As an addendum to the judgment, Master McQuail explained that the defendants were ordered to file and serve a bundle containing copies of reports of all cases referred to in the second statement (unless advocates agreed the cases were to be included in a common authorities bundle) and to be clear in the index whether any of the cases referred to do not exist. When filed, the bundle listed 6 cases as “No”.
The court then directed an explanation from the defendant by witness statement. In a witness statement by a paralegal at the defendant firm, it was explained that they had been responsible for preparing the managing partner’s witness statement and had used the AI overview section of Google during his research. This judgment noted that the Administrative Court had given a very clear warning about the misuse of AI and the misleading citation authorities in Ayinde.
Speculation of judicial use of AI
On the other side of the table, there has been speculation against the judiciary regarding use of AI in a judgment. The matter of Sandie Peggie v Fife Health Board [2025] 12 WLUK 249 concerned claims arising from the Equality Act 2010 being heard in the Employment Tribunal.
Whilst this is a case arising in the Scottish Employment Tribunal and the substance of this case does not relate to the use of AI, the errors in the Tribunal’s decision led to suggestions that the document had been generated with the assistance of AI. The errors, including a mis-quote from a previous judgment have been widely publicised. A first correction was issued in connection with this error on 11 December 205. Eleven further corrections to the decision were then issued on 23 December 2025. A formal complaint of judicial misconduct was made however it has most recently been reported that the president of the Employment Tribunals Scotland is satisfied the Judge did not use AI in drafting the judgment.
Lessons learned
It is clear from the growing case law in this area that legal practitioners are continuing to use AI in their work and are not always taking adequate steps to verify the material it generates. It is further evident that the courts consider it is necessary to maintain their hard-line approach to professionals who fall short of the required standards when using AI in legal research and drafting.
Recent case law reiterates to legal professionals that the potential serious consequences of failing to responsibly use AI include public admonishment, an adverse costs order (including a wasted costs order), referrals to professional bodies, cases being struck out, negligence claims, contempt proceedings or even a referral to the police.
As the use and applications of AI inevitably continue to grow, practitioners must ensure that they independently verify all references to case law and legal principles against authoritative sources and read the judgments so that they are used correctly. Ultimately, legal representatives remain responsible for the accuracy of materials submitted to the court regardless of how well known or trusted the software they are using is. Supervision and training of fee earners at all levels is also fundamental to the responsible use of AI. We are reminded by these cases that the technology we have available needs to augment, not replace, our own judgment and research.
Reproduced from Practical Law with the permission of the publishers. For further information visit www.practicallaw.com.