- within Litigation and Mediation & Arbitration topic(s)
- with Senior Company Executives, HR and Inhouse Counsel
- in United States
- with readers working within the Insurance, Media & Information and Securities & Investment industries
As I near the 35th anniversary of my Call to the Bar of Ontario, I struggle with the proper role that artificial intelligence should play in the legal research plan. As one of the lawyers in my firm responsible for teaching legal research to our summer and articling students, I strongly emphasize that our youngest lawyers should avoid the temptation of using AI to conduct legal research because of the mounting evidence that AI platforms are hallucinating cases and quotations and simply providing misleading answers. Studies suggest that even the AI products that are being developed and marketed by commercial legal publishers to assist in the conducting of legal research are not providing accurate answers. Although users of AI platforms are being warned to check and verify the answers produced by their AI research, this is clearly not always happening, even within well-renowned and sophisticated law firms who have adopted comprehensive AI policies.
In Cork v. Smith, [2026] EWHC 1199 (Ch), the law firm of Pinsent Masons LLP became the focus of the court’s attention as a result of a young lawyer’s use of and reliance upon AI that produced an incorrect answer to a legal problem.
Pinsent Masons LLP is an international law firm headquartered in London, England and is highly-regarded for its work.
In this case, the firm represented the applicants in an insolvency proceeding.
On March 17, 2026, a block transfer application was brought for the removal of the respondent from his office as administrator or liquidator in relation to a small number of administrations and voluntary liquidations. Under The Insolvency (England and Wales) Rules 2016 (the “IR”), a procedure permits insolvency practitioners, who often hold different types of appointment in respect of individuals and companies, to be removed from office as an office-holder of all of their appointments through a “block transfer application”. In other words, the court has the power under the rules to remove an insolvency practitioner from office as a trustee in bankruptcy, liquidator, administrator, or supervisor and to appoint another person in their stead.
The applicants’ law firm prepared a draft order and provided it to the court. Among other things, the draft order provided that the respondent would be released from liability in respect to each of his cases, without the need to apply to the Secretary of State.
Under England’s Insolvency Act, 1986 (the “Act”), the court explained that although this position is clear for administrators, it is less clear for liquidators.
Accordingly, on March 20, 2026, the court delivered a letter to Pinsent Mason LLP asking, “What power is there to order release?” The court noted that there was no express power in connection to a liquidation under sections 173 and 174 of the Act.
Ten days later, Pinsent Masons LLP responded. The response referenced a specific power under the IR to provide the release of a liquidator on a block transfer application. The language of this rule was presented as a quotation.
This surprised the judge who, after checking the IR rule cited by the law firm, concluded that no such power existed.
The judge was concerned that the IR rule cited to support the express power was an AI hallucination that arose from the law firm’s research. The judge brought this concern to the law firm’s attention in a subsequent letter dated April 14, 2026.
The law firm’s reply to this letter “astonished” the judge because even though it was acknowledged that a certain phrase identified by the judge was not in the rule, the law firm submitted that the wording connected to the rule as set out in their March 30 letter was a summary conclusion, drawn from a reading of several IR subrules. The law firm’s letter then proposed the same alternative course as set out in the March 30 letter.
The judge found this impossible to accept because the text as set out in the March 30 letter could not have been intended as a mere summary. Direct quotations had been used.
Accordingly, the law firm’s April 14th response caused the court to have further concerns about the accuracy of the legal position that was being taken, and the judge ordered the lawyers to, among other things, essentially explain why the March 30 letter had set out non-existing text and whether AI had been used.
The law firm admitted that AI had been used in the research and in the production of the two letters, and that the research had been conducted by a junior lawyer, who was only identified by the initials LA.
The evidence provided by the firm to the court revealed that LA had engaged in a 59-page chat with an AI tool when preparing the submissions and the letters.
A review of this chat showed that hallucinations began almost immediately.
Notwithstanding the hallucinations and repeated warnings from the AI tool to LA to verify the accuracy of its answers throughout the matter, LA never read an authoritative online resource or book or checked the AI’s answers. Rather, LA appeared to have entirely relied upon AI to provide the answers.
LA also did not appear to have advised other lawyers who worked on and had day-to-day responsibility for the client’s matter and who had approved and executed the letters to the court.
While the senior associate who had carriage of the matter and signed the letters was of the view that there was no express power granting the release of a liquidator, she simply believed that LA had found something that she had on a previous occasion been unable to find.
In her response to the court’s investigation into the submissions, she stated that she was “mortified” that inaccurate statements of the law had been provided to the court.
A partner, who also worked on the matter, simply approved the March 30 letter, which contained the hallucinated text. He too was unaware that LA had used AI in preparing that letter. The partner accepted that the April 14 response to the court also should not have been written in the manner as presented and stated that the firm was now putting further safeguards in place regarding the use of AI.
The firm also self-reported this matter to the professional regulator.
LA also emailed the professional regulator to disclose their identity.
In the circumstances, the court found that it was unnecessary to order contempt proceedings against any of the lawyers involved or to refer any of them to the professional regulator.
However, the court emphasized the dangers of using AI and that traditional legal research remains an important skill. The court concluded its judgment as follows:
I do not intend to set out the words of the President in Ayinde (R (Ayinde) v. London Borough of Haringey, [2025] EWHC 1383 (Admin)) again but it will be clear from what she said that legal professionals bear ultimate responsibility for their work and cannot outsource the process of legal research or of legal reasoning to an AI. It is a tool to be used with caution. AI has the potential to be wholly unreliable. AI may of course provide a jumping off point for research and legal reasoning but it does not, at least at present, do away with the need for proper research and thought on the part of a legal professional, even a very junior legal professional. (All emphasis added)
The key takeaway of this decision is that traditional legal research remains of utmost importance when answering legal issues. Junior lawyers, in particular, must take the time to understand traditional legal research methodology and appreciate the value of traditional secondary sources when answering legal issues. In my view, a junior lawyer who believes that AI platforms provide a “magic toolbox” of answers to any legal question or issue is a fool and should strongly be dissuaded from such a belief. Similarly, senior lawyers and partners should also avoid using AI platforms or programs to conduct legal research and should get into the habit of questioning their junior lawyers and students about whether they have used AI in the production of their work. The growth of AI tools and the temptation to use it in completing legal tasks requires everyone in the chain of a file to disclose, check and verify.
It is unknown if the professional regulator will heavily discipline or penalize the lawyers involved in this matter or penalize them at all. It is also unknown whether Pinsent Masons LLP will take any “disciplinary” action against their own lawyers. Although Pinsent Masons LLP and the individual lawyers named in the decision will suffer embarrassment and potentially some reputational loss, this will arguably be soon forgotten when the next lawyer or high-profile law firm is inevitably caught citing fake cases, statutory provisions or quotations in submissions to a court or tribunal. A PDF version is available to download here.
[This blog was not written with the assistance of artificial intelligence.]
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
[View Source]