Aug 2026
Executive Summary
In Anthony Malcolm Cork & Anor v Mark Smith [2026] EWHC 1199 (Ch), a junior associate (the “Associate”) from a reputable international firm (the “Firm”) put before the Court a fabricated version of Rule 12.37(5) of the Insolvency (England and Wales) Rules 2016 (“IR 12.37(5)”) in a block transfer application, seeking the release of the outgoing liquidator (the “Relief”). This confused the Court, with Judge Mullen (the “Judge”) expressing uncertainty as to what power allowed the Court to grant the Relief.
The Court emphasised and upheld professional standards when it was discovered that the Associate used AI to draft two letters to respond to the Judge’s legal queries regarding the false text of IR 12.37(5). The Court found that senior legal professionals at the Firm had failed to supervise the Associate adequately, which resulted in a non-existent statutory provision being put before the Court.
Factual Background
On 17 March 2026, the Firm filed a block transfer application (the “Application”). A block transfer application is an administrative court procedure made pursuant to Rules 12.36 and 12.37 of the Insolvency (England and Wales) Rules 2016 (“IR 2016”) to permit the removal and appointment of insolvency practitioners as officeholders through a single application. The Court has the power to remove and replace officeholders, including trustees in bankruptcy, liquidators, administrators, or supervisors of voluntary arrangements, by an uncontested application made on paper.
The Application submitted by the Firm sought two things: (i) removal and replacement of the officeholder and (ii) the release (or discharge) of the outgoing officeholder from liability. In relation to administrators, the position is clear: pursuant to paragraph 98 of Schedule B1 under the Insolvency Act 1986 (“IA 1986”), a person who ceases to be an administrator is discharged from liability at a time specified by the court. By contrast, under section 173 of the IA 1986, a liquidator’s release following removal by the court is determined by the Secretary of State, and there is no express power for the court to order release in these circumstances. Given the absence of any express provision in IR 2026 and IA 1986 authorising the Court to grant a liquidator’s release, the Judge wrote to the Firm seeking clarification of the basis upon which such Relief could be granted.
The Firm responded with a letter on 30 March 2026 (the “30 March Letter”) and set out IR 12.37(5) in italics (the “Purported Text”), claiming that IR 12.37(5) contained an express power to release liquidators, surprising the Judge who was unaware of any such procedure. After cross-checking the legislation and discovering that the statutory text bore no resemblance to the Purported Text in the letter, the Judge stated that his “first thought” was that the error may have been caused by an AI hallucination, which refers to the generation of inaccurate or fabricated information by AI. The Judge wrote back to the Firm and requested them to clarify the discrepancies between the Purported Text and the statutory legislation. He also enquired whether AI had been used in drafting the 30 March Letter, following which a further misleading letter dated 14 April 2026 was submitted to the Court (the “14 April Letter”).
Ultimately, three witness statements were filed. A partner and a senior associate of the Firm confirmed that they had failed to supervise the Associate adequately but stated that they were unaware at the time that AI was being used by the Associate. A third witness statement from the Deputy General Counsel (Risk & Regulatory) and the Compliance Officer for Legal Practice at the Firm, confirmed, by reference to the records of the Firm’s AI program, that the Associate had used AI in the preparation of both letters. Transcripts of the Associate’s chats with the AI, running to some 59 pages, were exhibited.
A letter was also received from a partner and chair of the Firm’s Innovation and Change Board, explaining the Firm’s AI pilot and detailing further safeguards being put in place.
The Firm agreed to pay the costs arising from the application to their now-former client, who were represented by new solicitors thereafter and resubmitted their block transfer order. The Firm also referred itself to the SRA. The reapplication was granted on 8 May 2026 by the same Judge.
Decision of the Court
The Court assessed whether to take further action against the solicitors involved in submitting two misleading letters to the Court, including whether contempt proceedings should be commenced. In doing so, the Court drew a distinction between the two letters: the 30 March Letter was primarily the product of an AI hallucination that had not been checked against authoritative sources, whereas the 14 April Letter was not primarily the result of an AI hallucination but was rather a construction, after the event, of a rationale for the 30 March Letter, which itself constituted a further instance of misleading information being placed before the Court.
The Court applied the factors identified in R (Ayinde) v London Borough of Haringey [2025] EWHC 1383 (Admin), including:
- the importance of setting and enforcing proper standards;
- the circumstances in which false material came to be put before the court;
- whether an immediate, full and truthful explanation is given to the court and to other parties to the case;
- the steps taken to mitigate the damage, if any;
- the time and expense incurred by other parties to the case, and the resources used by the court in addressing the matter;
- the impact on the underlying litigation; and
- the overriding objective of dealing with cases justly and at proportionate cost.
The Court found that both letters were misleading and that the Firm had failed to supervise the Associate adequately and to verify AI-generated material. Nevertheless, there was insufficient evidence that the senior solicitors intended to mislead the Court. The Court noted that the Firm had co-operated fully, disclosed the AI chat transcripts, accepted responsibility, compensated their clients for additional costs, reviewed its AI policies and voluntarily self-referred to the SRA.
Notably, the AI chat transcripts revealed that the AI tool itself had repeatedly warned the Associate to verify the statutory text against authoritative sources before filing. On multiple occasions, the AI tool expressly stated that it could not confirm the precise wording of IR 12.37(5) and urged the Associate to check it on legislation.gov.uk, cautioning that “the last thing you want is to cite a provision to the court with inaccurate wording”. Despite these warnings, the Associate did not appear to have checked the text. The Court described this omission as “inexcusable”.
Further, when the AI tool produced a draft of the 14 April Letter that included an “unreserved apology” to the Court, the Associate instructed the tool to remove it, stating: “I don’t think we should apologise – no”. The Court noted that this raised questions about the Associate’s judgment.
Although there were concerns regarding the conduct of the Associate, determining whether contempt had been committed would require further evidence, representations and additional court resources. It was determined disproportionate to pursue contempt proceedings where the conduct appeared to stem from “a serious lack of care and of judgment on the part of [the Associate] rather than a want of honesty”. Whilst the Judge described the Associate’s conduct as “very troubling”, he considered that the SRA was better placed to investigate the Associate’s conduct, particularly given the Associate’s junior status, the limited harm caused, and the additional court resources that further proceedings would require.
Instead, the Court concluded that public admonishment and Firm’s self-referral to the SRA was the appropriate and proportionate response.
Thoughts
The use of AI in legal practice is no longer a matter of choice but an inevitable reality. The speed of development in AI technology is, by any measure, unimaginable. Leading law firms across the globe are investing tremendous resources in AI development and integration, recognising that those who fail to adapt risk being left behind. Cork v Smith does not, and should not, stand as a cautionary tale against the use of AI in legal practice. Rather, it serves as a stark reminder that the purpose of AI must be to enhance the quality of legal work, not to replace the individual judgment of lawyers.
The case exposes a particular hazard in the use of AI by junior lawyers. A senior practitioner, confronted with an AI-generated provision that purported to quote a statutory rule, would instinctively reach for the statute itself. A junior lawyer, lacking the depth of experience to recognise what looks right and what does not, may be more susceptible to accepting AI output at face value. The very power of AI to produce fluent, plausible and authoritative-sounding text is precisely what makes it dangerous in the hands of those who do not yet have the professional instinct to question it.
Legal practice is, after all, a profession built on true professional judgment and personal accountability. No AI tool, however sophisticated, can absolve a solicitor of the duty to verify the accuracy of what is placed before the court. The temptation to rely on AI, particularly when it produces outputs that are coherent, well-structured and apparently convincing, must be tempered by the recognition that legal professionals bear ultimate responsibility for their work and cannot outsource the process of legal research or legal reasoning to AI, as Judge Mullen emphasised.
Cork v Smith also raises a question as to the effectiveness of standard AI use policies. The Firm had an AI policy in place at the material time, which expressly warned that AI outputs “may be inaccurate, biased or violate firm policies” and that users were “responsible and accountable for work produced with the assistance of AI”. The policy required that the “quality, accuracy and completeness of all client work produced with the assistance of AI must be checked in the usual way, by an appropriately qualified person”. Yet none of this prevented what occurred. A policy, however well drafted, is only as effective as the culture of compliance that supports it and the rigour with which it is enforced in day-to-day practice. Law firms must look beyond written policies to invest in meaningful training, active supervision and, critically, a culture in which junior lawyers feel empowered to acknowledge uncertainty and seek guidance rather than outsource their thinking to a machine.