Home » Blog » When ChatGPT Goes to Court: R (Ayinde) v London Borough of Haringey; Al-Haroun v Qatar National Bank QPSC

When ChatGPT Goes to Court: R (Ayinde) v London Borough of Haringey; Al-Haroun v Qatar National Bank QPSC

Authored By: Anastasiia Sosnovska

Southampton Solent University

  1. Case Citation and Basic Information

Full case name: The King (on the application of Frederick Ayinde) v London Borough of Haringey; and Hamad Al-Haroun v Qatar National Bank QPSC and QNB Capital LLC (joined cases).

Citation: [2025] EWHC 1040 (Admin) and [2025] EWHC 1383 (Admin); Case Nos: AC-2024-LON-003062 and CL-2024-000435

Court: High Court of Justice, King’s Bench Division, Divisional Court (exercising the court’s Hamid jurisdiction).

Date of decision: 6 June 2025 (judgment handed down; underlying Ayinde wasted costs decision by Ritchie J given 3 April 2025, reported at [2025] EWHC 1040 (Admin)).

Bench composition: Dame Victoria Sharp, President of the King’s Bench Division, and Mr. Justice Johnson.

  1. Introduction

This case considers the application of generative artificial intelligence (AI) in legal proceedings and overlooks  professional and ethical obligations owed by lawyers to the court in England and Wales. The joined proceedings, heard under Divisional Court’s Hamid jurisdiction, arose from two distinct cases in which lawyers relied on AI-generated legal research that contained fabricated case citations, invented questions, and a material misstatement of statute.

  1. Facts of the Case

The Ayinde case

The claimant (Frederick Ayinde) applied for judicial review of the refusal made by the defendant, the London Borough of Haringey, to provide an interim accommodation while the claimant was awaiting a statutory review of homelessness. Mr Ayinde was represented by Haringey Law Center, a legal aid charity. Victor Amadigwe is the Chief Executive and solicitor at Haringey Law Center and his paralegal is Sunnelah Hussain. Sarah Forey, pupil barrister, was instructed to settle the grounds for judicial review[1]. The grounds drawn up by Ms. Forey contained a misleading statement of section 188(3) of the Housing Act 1996 and referred to five non-existent authorities[2]. In her passages, she had also recited the neutral citation number of an unrelated case.

The defendant’s solicitor (Mr. Greenberg) was unable to locate the cases and continued to request Mr. Amadigwe for copies. Therefore, none were awarded. [3]The dispute in substance was resolved by the offer of accommodation by the defendant, but the London Borough of Haringey, nevertheless, applied for a wasted costs order. That application was heard by Ritchie J on 3 April 2025[4] , where Ms. Forey’s changing explanations were dismissed[5] , her conduct found to be improper, unreasonable and negligent[6] following Senior Court Act 1981[7],  and she and the Law Center were ordered to pay £2,000 each[8]. Moreover, the judge has also referred the matter to the regulators (Bar Standard Board and Solicitors Regulation Authority). [9]

The Al-Haroun case

Hamad Al-Haroun has filed a suit against Qatar National Bank and QNB Capital LLC, seeking £89.4m in damages for breach of a financing agreement. The claimant was represented by a solicitor, Abid Hussain of Primus Solicitors. In the application to set aside the striking out order made by the defendants, Mr Al-Haroun and Mr Hussain relied upon forty-five cited authorities in their witness statements filed. A judicial assistant reviewed them, and found that eighteen of the cases did not exist, and that most of the remaining cases were either misstated or irrelevant to the arguments presented [10]   Consequently, on May 9, 2025, Dias J rejected their application, expressed serious concerns, and transferred the case to the Hamid jurisdiction [11]

  1. Legal Issues

The Divisional Court determined the following fundamental questions:

  1. Do lawyers violate their ethical and professional obligations to the court when they use AI-generated legal research with fake citations, without independently checking that research against reliable sources?
  2. Did each lawyer’s actions meet the requirements to start contempt of court proceedings based on the facts of each case?
  3. What penalties, regulatory referrals, or other actions are suitable in cases where misleading material has been produced or is suspected of being produced?

        5. Arguments Presented

5.1 Submissions for the lawyers referred in the Ayinde case

Helen Evans KC, on behalf of Ms. Forey, argued that the contempt threshold was not met because her client was still a pupil that received inconsistent pupilage training,  who is working in a demanding environment, and , therefore, did not know the citations were false. The defense counsel has also mentioned the errors had little effect on the underlying outcome, and Ms. Forey had since shown judgment and apologized.[12] Andrew Edge, on behalf of Mr. Amadigwe and Ms. Hussain stated that the Law Center was an overworked charity, emphasising that Ms. Hussain, a supervised paralegal, was innocent. [13]

5.2 Submissions for the lawyers referred in the Al-Haroun case

David Lonsdale, speaking on behalf of Mr. Hussain and Primus Solicitors, acknowledged that the behavior in question was quite serious and openly stated that it “could not be worse”[14]. In mitigation, he said that Mr. Hussain was unaware that the quotations and citations were fictitious, and that the lawyer who had previously examined the material had recommended against pursuing the application,  but had failed to spot the fake citations.[15] It was argued that no additional judicial referral was required considering Mr. Hussain’s voluntary self-referral to the Solicitors Regulation Authority.

  1. Court’s Reasoning and Analysis

The Divisional Court considered professional guidance available on AI ‘hallucination’ including Bar Council guidance[16], SRA Risk Outlook[17] a blog by the Bar Standards Board[18] referring to the case of Mata v Avianca Inc[19], and judicial guidance on the use of AI[20].  The court made it clear that any legal professional performing AI-assisted research, or relying on research carried out by someone else (whether a trainee, pupil, paralegal lay client ), has a personal duty to verify the AI output against authoritative sources, such as the National Archives caselaw database or the official Law Reports[21]. The court  has also acknowledged degrees of culpability: negligent reliance on unverified AI output was professional misconduct but was not contempt, whereas intentionally putting before the court material which the individual recognises to be false, or untruthfully denying the use of AI in a signed witness statement, might be contempt[22],  requiring proof of knowledge of the falsity or lack of honest belief in truth, applying JSC BTA Bank v Ereshchenko.[23]

This paradigm was used to analyse Ms. Forey’s contradictory answers, which included a description of a “box of copies” and a subsequent claim that she had only abused the powers she had “intended” to rely on. The deliberate inclusion of fake citations or the employment of AI tools followed by a false denial in a declaration of truth are two possible scenarios that the court recognised as potentially constituting contempt. The President of the King’s Bench Division acknowledged the threshold of contempt [24]but the court refused to start proceedings due to unresolved concerns about her pupilage supervision, her current self-referral, Ritchie J’s public criticism, and her acute inexperience.[25]

In the Al-Haroun case, the court found that Mr. Al-Haroun’s errors do not absolve his legal representatives of responsibility to act with integrity. [26] But the court confirmed that Mr. Hussain had not known the citations were false, having been misled by relying, unusually, on research conducted by his own lay client. This was a serious but genuinely negligent failure to verify, insufficient to meet the contempt threshold.

  1. Judgment and Ratio Decidendi

The Decision

In the Ayinde case, Ms. Forey was referred to the Bar Standards Board [27] , in addition to the referral already made by Ritchie J. Mr. Amadigwe was referred to the SRA for his inadequate response upon notice despite the mitigating context of an over-stretched legal aid charity [28]However, Ms. Hussain was found to have done nothing wrong and to have acted properly throughout. [29]

In the Al-Haroun case, Mr. Hussain was referred to the Solicitors Regulation Authority for lamentable failure to check the accuracy of material that is put before court.  Mr. Al-Haroun itself was disregarded, as the focus of the courts is to look at the conduct of the lawyers themselves. [30] No referral was made in respect of counsel, and the contempt threshold was found not to be met. [31]The court has also directed a copy of the judgment to be sent to the Bar Council, the Law Society, and the Council of the Inns of Court, inviting urgent consideration of further regulatory steps.[32]

Ratio Decidendi

Lawyers who use generative artificial intelligence (AI) tools to carry out legal research, and lawyers who rely upon legal research carried out by others, whether those others be lay clients, trainee solicitors/pupil barristers, are subject to an independent professional and ethical duty to check the accuracy of any such legal research against the most authoritative primary source(s) prior to presenting same to a court. The failure to do this would at least constitute professional negligence and/or a breach of the duty not to mislead the court; and possibly even contempt of court depending on the lawyer’s knowledge. The obligation to verify cannot be fulfilled merely by relying upon a client’s assurance that the research has been accurately conducted, nor an inexperienced junior colleague’s assurance that he/she has checked the research for accuracy, nor the AI tool’s own assurance that it has performed its research correctly.

  1. Critical Analysis

8.1 Significance of the Decision

This case draws together previously issued regulatory guidance from the Bar Council, the Solicitors Regulation Authority, and the judiciary, converting non-binding material into a judicially- established standard of conduct. The judgement also draws on comparable decisions such as Harber v HMRC [33]in the UK Tax Tribunal and Mata v Avianca Inc[34] in the United States, pointing to a broader international compliance on independent verification of AI-generated legal material.

8.2 Implications and Impact

The decision has immediate practical effect on legal practice. Professional bodies started to strengthen training and supervision[35], particularly of pupils, trainees, and paralegals working in under-resourced practices. The Haringey Law Centre context imposes a real access-to-justice question, since legal aid charities serving the most vulnerable clients are often the least equipped to provide real supervision. [36]Firms and chambers are now expected to implement clear verification protocols, and the judgment has already generated a developing line of authority, being applied in subsequent decisions such as Cork v Smith[37] and further tribunal cases addressing AI-generated material.

8.3 Critical Evaluation

The court’s approach in this case has real strengths. By distinguishing negligent reliance from dishonest one, it avoids an escalation of every instance of AI misuse into contempt proceedings. It recognises systemic under-resourcing and reflects a fair and realistic assessment of how AI misuse tends to arise in practice. However, the judgment leaves an important factual question formally unresolved: the court declined to determine, on the merits, whether Ms. Forey in fact used AI at all, notwithstanding finding the threshold for contempt met on either of two possible explanations. Such an ambiguity arguably weakens the doctrinal clarity of the ratio for future courts applying it. Further, the judgment does not prescribe concrete institutional verification standards, instead inviting professional bodies to consider further steps. This is a deferential approach that may result in inconsistent implementation across chambers and firms.

  1. Conclusion

This case establishes the foundational English authority on the professional consequences of relying on AI-generated legal research without independent verification. By including existing regulatory guidance into a binding judicial standard, and by carefully distinguishing negligent from dishonest conduct, the Divisional Court struck a proportionate balance between deterrence and fairness to inexperienced or under-resourced practitioners. The single most important takeaway is that the duty to verify AI-generated material against authoritative sources is non-delegable: it cannot be discharged by relying on a client, a junior colleague, or the reassurance of the AI tool itself. The decision leaves open how professional bodies will translate this standard into concrete supervisory and training requirements, and further litigation is likely to continue refining the line between excusable negligence and sanctionable misconduct as AI use in legal practice becomes increasingly apparent.

  1. Reference(S):

TABLE OF CASES

Cork and another v Smith [2026] EWHC 1199 (Ch)

Harber v Commissioners for His Majesty’s Revenue and Customs [2023] UKFTT 1007 (TC)

JSC BTA Bank v Ereshchenko [2013] EWCA Civ 829

Mata v Avianca Inc, 678 F Supp 3d 443 (SDNY 2023)

R (Ayinde) v London Borough of Haringey [2025] EWHC 1040 (Admin)

R (Ayinde) v London Borough of Haringey; Al-Haroun v Qatar National Bank QPSC [2025] EWHC 1383 (Admin)

TABLE OF LEGISLATION

Housing Act 1996, s 188(3)

Senior Courts Act 1981, s 51(7)(a)

BIBLIOGRAPHY

Bar Council, ‘Considerations when Using ChatGPT and Generative Artificial Intelligence Software Based on Large Language Models’ (30 January 2024) https://www.barcouncil.org.uk/resource/new-guidance-on-generative-ai-for-the-bar.html accessed 31 August 2026

Bar Standards Board, ‘ChatGPT in the Courts: Safely and Effectively Navigating AI in Legal Practice’ (8 October 2023) https://www.barstandardsboard.org.uk/resources/chatgpt-in-the-courts-safely-and-effectively-navigating-ai-in-legal-practice.html accessed 31 August 2026

Bar Standards Board, ‘New Guidance Supports Barristers to Safely Adopt Artificial Intelligence and Emerging Technologies’ (18 May 2026) https://www.barstandardsboard.org.uk/resources/new-guidance-supports-barristers-to-safely-adopt-artificial-intelligence-and-emerging-technologies.html accessed 31 August 2026

Courts and Tribunals Judiciary, ‘Artificial Intelligence (AI): Guidance for Judicial Office Holders’ (first published December 2023, updated 14 April 2025) https://www.judiciary.uk/wp-content/uploads/2025/04/Refreshed-AI-Guidance-published-version-website-version.pdf accessed 31 August 2026

Solicitors Regulation Authority, ‘Risk Outlook Report: The Use of Artificial Intelligence in the Legal Market’ (20 November 2023) https://www.sra.org.uk/sra/research-publications/artificial-intelligence-legal-market/ 

UK Parliament ‘Written evidence submitted by Young Legal Aid Lawyers (YLAL) (LAI 0033)’ (September 2013) https://committees.parliament.uk/writtenevidence/45816/pdf/

[1] R (on the application of Frederick Ayinde) v The London Borough of Haringey; Hamad Al-Haroun v Qatar National Bank Qpsc, QNB Capital LLC [2025] EWHC 1383 (Admin) (Dame Victoria Sharp) [32]

[2] Ibid, [38]

[3] Ibid, [39]

[4] [2025] EWHC 1040 (Admin)

[5] Ibid (n3)  (Ritchie J) [53]

[6] Ibid,(n3)  [64]

[7] Senior Courts Act 1981, s51(7)(a)

[8] Ibid(n3), [72]

[9] Ibid (n8)

[10] Ibid(n1), [74]

[11] Ibid(n1), [73]

[12] Ibid(n1), [60]

[13] Ibid(n1), [61]

[14] Ibid(n1), [78]

[15] Ibid(n14)

[16] The Bar Council ‘Consideration when using ChatGPT and generative artificial intelligence software based on large language models ‘(30 January 2024)https://www.barcouncil.org.uk/resource/new-guidance-on-generative-ai-for-the-bar.html [17]

[17] Solicitors Regulation Authority, ‘Risk Outlook report; the use of artificial intelligence in the legal market’ (20 November 2023) https://www.sra.org.uk/sra/research-publications/artificial-intelligence-legal-market/

[18] Bar Standards Board ‘ChatGPT in the Courts: Safely and Effectively Navigating AI in Legal Practice’ (8 October 2023) https://www.barstandardsboard.org.uk/resources/chatgpt-in-the-courts-safely-and-effectively-navigating-ai-in-legal-practice.html

[19] 678 F Supp 3d 443 (SDNY 2023)

[20] Courts and Tribunals Judiciary ‘Artificial Intelligence (AI); Guidance for Judicial Office Holders’ ( first provided in December 2023 and updated 14 April 2025) https://www.judiciary.uk/wp-content/uploads/2025/04/Refreshed-AI-Guidance-published-version-website-version.pdf

[21] Ibid(n1), [7]

[22] Ibid(n1), [26]

[23] [2013] EWCA Civ 829

[24] Ibid(n1), [68]

[25] Ibid(n1), [69]

[26] Ibid(n1), [81]

[27] Ibid(n1), [70]

[28] Ibid(n1), [72]

[29] Ibid(n1), [73]

[30] Ibid(n1) , [79]

[31] Ibid(n1), [80]

[32] Ibid(n1), [82]

[33] [2023] UKFTT 1007 (TC)

[34] Ibid(n19)

[35]Bar Standard Board ‘New Guidance supports barristers to safely adopt Artificial Intelligence and emerging technologies’ (18 May 2026) https://www.barstandardsboard.org.uk/resources/new-guidance-supports-barristers-to-safely-adopt-artificial-intelligence-and-emerging-technologies.html

[36] UK Parliament ‘Written evidence submitted by Young Legal Aid Lawyers (YLAL) (LAI 0033)’ (September 2013) https://committees.parliament.uk/writtenevidence/45816/pdf/

[37] [2026] EWHC 1199 (Ch)

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top