Briefing

The use of AI in preparing court documents: Why the Civil Justice Council’s review matters

Introduction

The use of artificial intelligence (AI) in legal practice has moved rapidly from experimentation to everyday reality. Tools capable of drafting pleadings, summarising evidence, and proposing legal arguments are now widely accessible to legal practitioners at all levels. Against this backdrop, the Civil Justice Council (CJC) established a Working Group in January 2025 on the use of AI in preparing court documents, and has now launched a Consultation, which will inform its review.

The Working Group, chaired by Lord Justice Birss, Deputy Head of Civil Justice, will consider whether specific rules are required to govern the use of AI by legal representatives when preparing court documents.

The consultation will close on 14 April 2026.

This article explores the need for regulatory oversight, the risks and challenges of AI use in litigation, and offers commentary on AI adoption in English litigation.

Why is this review needed?

AI Is Already Being Used in Litigation Practice

AI tools are already being used when:

drafting statements of case;
for disclosure;
generating chronologies;
drafting legal arguments or finding authorities; and
assisting with case strategy.

This use has largely developed without formal regulatory guidance, leaving practitioners to rely on professional judgment, existing duties, and vendor assurances. The absence of clear standards creates uncertainty, inconsistency in practice, and a high risk of error.

Integrity of The Civil Justice System

The integrity of civil litigation depends heavily on the evidence and documents placed before the court: statements of case, witness statements, and expert reports are important documents on which both the judge and opposing parties rely.

One of the CJC’s statutory functions is to ensure the civil justice system remains accessible, fair, and efficient. However, unregulated or poorly understood use of AI could undermine all three:

fairness, if inaccuracies go undetected;
accessibility, if there is disparity in the use or access to AI tools by parties; and
efficiency, if AI‑generated errors lead to further disputes or increased time and costs.

Using AI in litigation – the key issues:

1. Accuracy and “Hallucinations”

One of the most widely discussed risks of generative AI is its tendency to produce content that appears real but is actually factually or legally incorrect – commonly referred to as ‘hallucinations’. In a litigation context, this may include:

fictitious case law;
incorrect quotations from cases;
misstatements of procedural rules; or
flawed summaries of evidence.

Courts in several jurisdictions have already criticised or sanctioned lawyers for filing documents containing AI‑generated false material. For example, in R (on the application of Ayinde) v London Borough of Haringey [2025] EWHC 1040 (Admin), the defendant’s team referenced five AI-hallucinated cases and were eventually referred to the Bar Standards Board and the Solicitors Regulation Authority. Meanwhile, in Bandla v SRA [2025] EWHC 1167 (Admin), the inclusion of AI-hallucinated case law in an appeal resulted in the case being struck out as an abuse of process.

2. Accountability and Professional Duties

AI does not owe duties to the court. Legal representatives do. Legal practitioners are subject to a number of core duties, including the duty to act with honesty, integrity, independence, and to act in a way that upholds public trust and confidence. These duties cannot be delegated to generative AI – no matter how advanced it may seem. Where AI contributes to, or produces a court document, questions arise as to:

who is actually responsible for errors;
what level of human verification is required; and
whether existing professional conduct rules are sufficient.

3. Witness Statements: A Particular Risk Area

Witness statements are intended to reflect the witness’ own recollection, expressed in their own words. The use of AI to draft or heavily edit witness evidence risks distorting genuine recollection, introducing language or structure alien to the witness – all of which may ultimately damage the witness’ credibility at trial.

Judicial concern about over‑lawyered witness statements predates AI; however, the introduction of generative tools risks intensifying this existing problem.

4. Expert Evidence and Independence

Expert reports present similar concerns. Experts owe overriding duties to the court and AI assistance in drafting or analysis, particularly if undisclosed, risks compromising the perception of an expert’s independence, obscuring the expert’s true methodology, and introducing material that falls outside of the expert’s area of expertise.

The case for guardrails, not prohibition

AI as a Legitimate Efficiency Tool

Used appropriately, AI can reduce the time associated with repetitive drafting, improve access to justice by lowering costs, assist with early case assessment, and allow lawyers to focus on strategic and advisory work – rather than on administrative tasks.

An outright ban would be both unrealistic and counterproductive. The question should not be whether AI should be used, but how.

A Risk‑Based Approach

A sensible framework should distinguish between:

low‑risk uses (e.g. formatting, summarising disclosed documents, generating first drafts of pleadings subject to full review); and
high‑risk uses (e.g. drafting witness statements, expert opinions, or documents requiring personal knowledge or independent judgment).

This approach aligns with how courts already treat different categories of evidence.

HFW comment

If AI is to be embraced within English civil litigation, certain core principles must be clearly articulated. The key principle being that human responsibility cannot be displaced. All court documents prepared with AI assistance should be subject to human verification in the same way the work of a junior lawyer would be reviewed by their supervisor, with the legal representatives retaining full accountability for accuracy and compliance with procedural rules. Where AI has been used significantly when producing a document, transparency should be expected – not to deter innovation, but to preserve trust in the Civil Justice System.

Any regulatory response should adopt a flexible, principles‑based approach that accommodates technological evolution while safeguarding the integrity of the Civil Justice System. The credibility of the Civil Justice system depends on courts being able to trust what is placed before them. AI can support that process – but only if it is used responsibly, transparently, and with a clear understanding of its limits.

We encourage all involved in the civil litigation process to provide feedback on the Consultation, and we will report further once the findings are known.

The link to the Consultation is here.

With thanks to Danica Douglas, Paralegal, for her assistance in producing this article.