Original essay question / assignment brief
The Legal Framework of Responsibility for Self-Represented Litigants Submitting Claims with AI-Generated Errors in the UK and Implications on Access to Justice and Procedural Fairness
Introduction
The justice system in England and Wales is witnessing a significant increase in self-represented litigants (SRLs), often termed litigants in person. This trend, driven by factors such as reductions in legal aid, places considerable strain on both the individuals navigating a complex legal system and the courts themselves. Concurrently, the rise of accessible generative Artificial Intelligence (AI) platforms like ChatGPT offers a seemingly powerful tool for SRLs to draft documents and research law. However, this technology carries substantial risks, notably the potential to generate plausible but erroneous or entirely fabricated information, including fake case law. This essay will examine the legal framework governing the responsibility of SRLs who submit court documents containing such AI-generated errors. It will argue that while the existing procedural rules place ultimate responsibility on the litigant, the courts face a difficult balancing act. They must uphold procedural integrity without disproportionately penalising vulnerable litigants, thereby ensuring the concurrent, and sometimes conflicting, principles of access to justice and procedural fairness are maintained.
The Existing Procedural Framework for Litigant Errors
The conduct of civil litigation in England and Wales is governed by the Civil Procedure Rules (CPR). The CPR's overriding objective is to enable the court to deal with cases justly and at proportionate cost (CPR 1.1). This includes ensuring parties are on an equal footing and saving expense. While courts are expected to provide some measure of assistance to SRLs, this does not absolve them of their duty to comply with the rules. The Supreme Court in Barton v Wright Hassall LLP [2018] UKSC 12 confirmed that being an SRL is not a reason to be excused from compliance with rules and court orders, though it may be a relevant factor for the court to consider.
Under the CPR, the court possesses robust case management powers to deal with deficient claims. For instance, CPR 3.4 allows a court to strike out a statement of case if it appears to be an abuse of the court’s process or if it discloses no reasonable grounds for bringing the claim. A claim reliant on non-existent legal authority would almost certainly fall into one of these categories.
Furthermore, a critical element of responsibility is the statement of truth. CPR 22.1 requires that certain documents, including statements of case, be verified by a statement of truth. By signing this, the party confirms their belief that the facts stated in the document are true. If a party submits a document containing information they know to be false, or without an honest belief in its truth, they may be found in contempt of court (CPR 32.14). While an SRL using AI may not know a cited case is fake, their duty is to have an honest belief in the contents of their submission. This places a burden of verification on the litigant, regardless of how the document was generated.
AI-Generated Errors and the Judicial Response
The unique danger of generative AI in a legal context is its propensity for ‘hallucination’ – the invention of facts or sources that appear credible. This was starkly illustrated in the US case of Mata v Avianca, Inc. (2023), where lawyers were sanctioned for submitting a brief containing multiple fictitious case citations created by ChatGPT. While not binding in the UK, this case served as an international warning of the technology's pitfalls.
The first significant judicial consideration of this issue in the UK arose in the First-tier Tribunal case of Harber v HMRC [2023] UKFTT 1007 (TC). In this tax case, the applicant, an SRL, used ChatGPT to find supporting case law and subsequently included nine fictitious cases in their skeleton argument. The Tribunal noted the "wasted time and expense" for HMRC in identifying the errors and the "disruption to the tribunal" (para 39). While the Tribunal decided not to take further action against the litigant, citing her lack of intent to deceive, it issued a clear warning about the risks of using AI without independent verification. Judge Sinfield stated, "checking the accuracy and reliability of AI-generated content is essential" and that "litigants who use AI to generate submissions… will be held responsible for their content" (para 41).
This case indicates the likely judicial approach: while some initial leniency may be shown to an SRL who has made an honest mistake, the fundamental responsibility for the accuracy of submissions remains with the person filing them.
The Locus of Responsibility and Emerging Guidance
The principle established in Harber is that the user, not the AI developer, is responsible for the output. This aligns with the function of the statement of truth and the general duties owed to the court. The litigant is the one making the representations to the court and the opposing party, and they cannot delegate their duty of candour and accuracy to a machine.
This position has been reinforced by senior judicial figures. In December 2023, guidance issued by HHJ Bird for the Civil and Family Courts in Manchester stated that AI could be used for case summaries or document drafting, but with the "important caveat that the party or their legal representative is responsible for the content" (para 4). More authoritatively, the Master of the Rolls, Sir Geoffrey Vos, has stated that while AI is a tool to be embraced, "litigants and lawyers must remain responsible for the material they put before the court" (Vos, 2024). This emerging consensus from the judiciary signals a clear direction: AI is a tool, and like any other research tool, its output must be checked by the human user before it is submitted to a court. An SRL who fails to do so runs the risk of their claim being struck out, facing adverse costs orders, or in egregious cases, contempt of court proceedings.
Implications for Access to Justice and Procedural Fairness
The strict allocation of responsibility to the SRL creates a significant tension with the principle of access to justice. For many, AI represents a vital resource that can help level the playing field against represented opponents. It can help demystify legal jargon, structure arguments, and suggest avenues of research, thereby empowering SRLs to present their cases more effectively (Susskind, 2023). If the courts adopt an overly punitive approach to AI-generated errors, it may deter SRLs from using these potentially beneficial tools or, worse, from pursuing valid claims altogether. This would create a new barrier to justice, ironically stemming from a technology that promises to enhance it.
Simultaneously, the courts must uphold procedural fairness for all parties. A claim based on fabricated legal authority is manifestly unfair to the responding party, who must expend time and legal costs to identify and rebut the non-existent law. This diverts resources and undermines the overriding objective of dealing with cases at a proportionate cost. It also represents a significant abuse of the court’s process, consuming precious judicial time and undermining the integrity of the proceedings.
Therefore, the courts must navigate a careful path. A proportionate response might involve distinguishing between a genuine, one-off mistake by a vulnerable SRL, as seen in Harber, and a persistent or reckless reliance on unverified AI output. In the former case, judicial guidance and an opportunity to amend the claim might be appropriate. In the latter, more robust sanctions like strike-out or costs orders would be necessary to protect the other party and the court process.
Conclusion
The legal framework in England and Wales places responsibility for AI-generated errors squarely on the self-represented litigant who submits them. The combination of the CPR's requirements, particularly the statement of truth, and emerging judicial guidance confirms that AI is merely a tool, and its user is accountable for its output. While the decision in Harber v HMRC demonstrates a degree of initial judicial understanding, it also serves as a firm warning that such leniency is not guaranteed.
The core challenge for the justice system is to reconcile this necessary allocation of responsibility with the broader goals of access to justice and procedural fairness. The potential for AI to assist SRLs is immense, but so are the risks it poses to the integrity of legal proceedings. A balanced approach is required, one that encourages the responsible use of technology while imposing appropriate sanctions for its misuse. As AI becomes more integrated into society, the courts will need to develop a consistent and proportionate jurisprudence that upholds the rule of law without closing the courthouse doors to those who are simply trying to navigate it without a lawyer.
References
Barton v Wright Hassall LLP [2018] UKSC 12.
Civil Procedure Rules 1998.
Harber v The Commissioners for His Majesty's Revenue and Customs [2023] UKFTT 1007 (TC).
Mata v Avianca, Inc., No. 22-CV-1461 (PKC) (S.D.N.Y. June 22, 2023).
Susskind, R. (2023) Tomorrow's Lawyers: An Introduction to Your Future. 3rd edn. Oxford University Press.
The Courts and Tribunals Judiciary (2023) <a href="https://www.judiciary.uk/wp-content/uploads/2024/01/AI-Guidance-in-the-Civil-and-Family-Courts-in-Manchester-final-211223.pdf">Practice Guidance on the Use of Artificial Intelligence in the Civil and Family Courts in Manchester</a>.
Vos, G. (2024) <a href="https://www.judiciary.uk/speech-by-the-master-of-the-rolls-sir-geoffrey-vos-ai-and-the-law-the-next-25-years/">AI and the law: The next 25 years</a>. Speech delivered at the University of Oxford.
