By Dr Sue Kee, Dr Brett Davies and Gad Coffie; members of the Ethics Committee of the Law Society
AI misuse cases continue to disrupt the Courts, but it is not just the legal practitioners who bear the consequences. Although some websites that track legal decisions on AI misuse have sprung up, we have not cited them as there is a risk that by the nature of their content, including use of AI to capture data, AI hallucinations may attach to those sources.
With updated Supreme Court AI guidances recently issued in Victoria, Chief Justice Niall observed, in relation to the Courts, across State and Federal jurisdictions, having variously approached the adoption and use of AI:
[f]or the most part, they are reasonably consistent in their approach – non-prescriptive, permissive, cautionary…Looking ahead, national consistency is desirable. In practice, I expect this will be an organic process: Remarks at The Victorian Justice Leaders AI Forum, Melbourne, 26 May 2026, p 6.
Notably, the Chief Justice clearly articulated that AI is not a substitute for human judgement. Honesty and trustworthiness are vital to community confidence in the legal profession: Australian Solicitors’ Conduct Rules 2022 with Commentary (2024) (ASCRs). This piece considers the contemporary challenge to upholding these values in a landscape awash with AI technologies vying for business uptake in a competitive legal market.
The paramount duty to the Court and the administration of justice is sourced in the common law: so much is reflected in Rule 3 of the ASCRs. The introductory commentary to the ASCRs gives expression to its long-standing public interest rationale, by reference to remarks made by then Supreme Court of Victoria Chief Justice, Marilyn Warren AC:
…It is the lawyers, through the duty owed to the court, who form the legal profession and who underpin the third arm of government, the judiciary. Without the lawyers to bring the cases before the courts, who would protect the citizen? Who would enforce the law? It is this inherent characteristic of the duty to the court that distinguishes the legal profession from all other professions and trades.
Mata v Avianca – a lesson in AI hallucinations
Mata v Avianca Inc, 678 F Supp 3d 443 (S.D.N.Y., 2023), was the ‘landmark US case revolving around the misuse of AI’: James Curlin IV, ‘ChatGPT Didn’t Write This…Or Did It?…’ (2025) 78 Arkansas Law Review 123, 125. Two litigators, each with decades of experience, together with their firm, were sanctioned by the Court, as they had:
…abandoned their responsibilities when they submitted non-existent judicial opinions with fake quotes and citations created by…ChatGPT, then continued to stand by the fake opinions after judicial orders called their existence into question.
To contextualise the firm’s technology, the lawyers both gave evidence that they had not used Westlaw or LexisNexis for research since law school. While it may be human instinct to avoid answering for mistakes more conveniently attributed to a machine’s algorithm, Mata aptly demonstrated this is erroneous. Lawyers must discharge the foundational ethical duty to the Court with courage, and with the conviction of candour, in all instances.
The following harms that flow from misplaced reliance on AI, set out in Mata, bear repeating:
- The opposing party wastes time and money exposing the deception.
- The Court’s time is taken from other important endeavours.
- The client may be deprived of arguments based on authentic judicial precedents.
- There is potential harm to the reputation of judges and courts whose names are falsely invoked as authors of the bogus opinions and to the reputation of a party attributed with fictional conduct.
- It promotes cynicism of the legal profession and the American judicial system.
- A future litigant may be tempted to defy a judicial ruling by disingenuously claiming doubt about its authenticity.
As AI becomes even more capable, it is critical that members of the legal profession do not lose sight of their ethical obligations. For example, Anthropic has recently reported, against the backdrop of an IPO, that its Claude AI is being progressed towards ‘recursive self-improvement’ capabilities: Aisha Down, “Anthropic Says The World Should Have Option To ‘Pause’ on AI”, The Guardian (Online, 6 June 2026).
Re an Office-Holder; Cork v Smith – an AI tale of missed opportunity
Fast forwarding three years to the present is the UK High Court of Justice case of Re an Office-Holder; Cork v Smith [2026] EWHC 1199 (Ch). ICC Judge Mullen held over an application that, in the ordinary course, was ‘administrative and uncontentious’; part of the ‘daily diet of electronic boxwork’ [10]-[11]. Nevertheless, the existence of a differentiation in the applicable statutory regime, as to the Court’s power to grant a liquidator a form of release from liability, available to an administrator upon removal from a case, was agitated [14]. So much was familiar to the Judge.
Perusal of the application, and orders sought, revealed an absence of a reference to such a distinction, purportedly supported by relevant statute. An explanation was sought from the applicant’s award-winning internationalised law firm [17].
Unpersuaded by the response, the Judge undertook his own necessary research; AI hallucination was suspected [20]. Further correspondence was exchanged: the Judge found the explanation proffered ‘impossible to accept’; indeed, His Honour was ‘astonished’ by it [23]. The Judge ordered the writer of the two letters submitted to the Court to file a witness statement, and for the responsible partner to file a confirming witness statement.
It transpired that a fairly recently admitted junior lawyer, supervised by a senior associate, had day to day conduct of the matter. The junior had used AI to prepare the application, as well as the correspondence to the Court [27]. His Honour said that where the AI referenced certain provisions in the statute, ‘on a large number of occasions, it is plainly wrong or, at the very least, extremely misleading’ [33].
The evidence revealed supervisory shortfalls, and gaps in internal communications as to AI use; at the material time, the use of AI was not disclosed by the junior lawyer to the senior associate, or the responsible partner. A focus on the communications within the law firm is illuminating. Using Teams for communication, the junior was instructed by the senior associate to:
…draft a letter to the court and “set out the Rule you’ve found as the rationale for us asking for release” [43].
An email to the client was also prepared. The documents drafted, using AI, were settled by the responsible partner as ‘good to go’, and reviewed by the senior associate and confirmed as ‘fine’ [45]. Curiously, the junior thereafter interrogated the AI which issued a warning to manually check the statute wording: the AI had not been able to verify this itself. The AI then offered to do this: ‘I would not want you to file without confirming the precise text from the primary source first’ [47]. The AI explained that it had used the instructor’s own wording, a paraphrase of the statutory provisions, which had then been elevated to italicisation (and thereafter treated as a citation, albeit not humanly verified).
Despite the AI issuing repeated warnings that the text of the statute needed to be checked, the opportunity for human intervention was squandered. The first explanatory letter, sent in the form that it was, resulted in the Court issuing the second request for clarification. For brevity, the factual matrix of the second letter is not repeated; it too was infected by poor judgement. The Judge considered it to be ‘an unjustifiable response and an opportunity to set the record straight’ [for an ‘egregious misstatement’] ‘became a further instance of misleading information being put before the court’ [72], [74].
The red flag – the essentiality of human judgement
The transcript of the communications between the junior and the AI tool are set out in some detail by Judge Mullins. The warnings generated by the AI were unambiguous red flags:
[40]…However, I want to be candid with you — I am not fully confident that I am reproducing the exact statutory wording of Rule 12.37(5) with complete precision. The substance of the provision is as I have described in our earlier discussion, but for a submission to the court you should verify the exact wording against the current version of the Insolvency (England and Wales) Rules 2016 as published on legislation.gov.uk before relying on it. The last thing you want is to cite a provision to the court with inaccurate wording.
Would you like me to assist with anything further on this point once you have verified the exact text?
Notwithstanding the unequivocal gap in the AI’s verification of the text of the statutory provision critical to the Court’s enquiry, and the emphasis (of the AI) on the importance of accuracy insofar as the work was being produced to the Court, the junior lawyer did not check the text. The opportunity to do so was wide open, but it was missed.
The firm referred itself to the Solicitors’ Regulation Authority (SRA), indicating that full cooperation would be given; a response considered appropriate by His Honour [82]. The firm chose to relieve the junior from putting on evidence [28] and so was not named in the public admonishment [69]. His Honour was, however, satisfied the SRA was aware of the involvement of the junior practitioner [92].
Costs were also borne by the firm, including for the alternative solicitors appointed to act for the applicant. The firm’s Deputy General Counsel (Risk & Regulatory) and Compliance Officer for Legal Practice submitted detailed evidence by way of the AI transcripts, as well as the firm’s AI use policy [27]. The Court also received an apologetic letter from the chair of the Innovation and Change Board explaining ‘the firm’s AI pilot, and the further steps taken to put safeguards on its use in some detail’ [29].
The case illustrates, again, the potential harms articulated in Mata three years previously. The sophistication of the AI tool, in this instance, demonstrates rapid technological advances. However, human judgement must be exercised by lawyers consistently with the paramount ethical duty. It is no answer that the discharge of the duty is unrealisable, cognitively, in the context of AI: see discussion on potential hazards of worker compensation claims, arising in connection with so called ‘AI brain fry’: David Marin-Guzman, “‘AI Brain Fry’ The New Frontier For Compensation Claims”, Australian Financial Review (Online, 31 May 2026). Nevertheless, the challenge of finding the balance is highlighted by former Australian Human Rights Commissioner, Ed Santow, now director at the University of Technology Sydney’s Human Technology Institute:
…research was finding AI was reducing workers’ capacity to do normal parts of their job, which created its own stressor. There are many more tasks or activities where essentially we’re outsourcing the thinking work…From that, two consequences are arising. You understand the issue less well. You get a response but if someone asks you to explain your response or come to it from a different angle you become less able to do that. You lose the knowledge and you lose the skill. Then people feel stressed in performing tasks that are core to what you’re employed to do…It’s not inevitable when using AI, but it’s a real problem – and that problem hadn’t arisen before we started using AI: Marin-Guzman.
Conclusions
In Mata a practitioner, among other things, misplaced his professional trust in a colleague, and the colleague misplaced his ethical duty with an unfettered reliance on AI to perform his legal research. Accuracy and authenticity of legal reasoning fell by the wayside, honesty and candour blurred. In Re an Office-Holder; Cork v Smith the misplaced trust reposed in the junior lawyer, alternatively a shortfall in supervision, is telling.
The supervising senior associate accepted the junior had identified, in the statute, a provision that supported the existence of an express power not known to the more experienced practitioner. The case reinforces the concluding remarks of Judge Mullins:
Legal professionals bear ultimate responsibility for their work and cannot outsource the process of legal research or of legal reasoning to an AI [95].
For a full appreciation of the ethical lesson to be drawn from this case, we recommend practitioners take note of the citation at [66], to President Dame Victoria Sharp, on the Court’s approach to the standard of ethical conduct in use of AI, in R (Ayinde) v London Borough of Haringey [2025] EWHC 1383 (Admin); [2025] 1WLR 5147, [4]-[9].
While there are jurisdictional differences, in an internationalised legal environment, honesty and candour are universal hallmarks of trust in judgement exercised consistently with the ethical obligation to discharge the duty to the Court and administration of justice.
The authors endorse the cautious optimism, expressed by Niall CJ at the Victorian Justice Leaders AI Forum:
…the adoption of AI can make the profession stronger. That requires a clear-eyed focus on those attributes of the legal profession that are unique and important in the administration of justice, p10.
Against this backdrop, we close with the observation of New South Wales Supreme Court Chief Justice, Andrew Bell AC, that lawyers operate in an AI technological growth environment that is ‘non-linear, apparently unstoppable and ever increasing in its speed and intensity’: Chief Justice Andrew Bell AC, ‘Corporate Responsibility and Directors’ Duties In The Era of Artificial Intelligence’ (2026 Harold Ford Memorial Lecture, University of Melbourne 21 May 2026), [6].
The authors are grateful to Paul D Evans for bringing the Re an Office-Holder; Cork v Smith decision to the attention of the members of the Ethics Committee.
