Given the proliferation of AI tools such as ChatGPT and Claude, it is unsurprising that their use in litigation has begun to feature in a number of judgments and has drawn the attention of the senior judiciary.
In this article, Anson Cheung and Ryan Fincham discuss recent judgments considering the use of AI in witness statements and analyse potential changes to the rules of the court to address the use of AI by legal professionals when preparing witness statements.
Recent Cases
The Civil Procedure Rules (“CPR”) clearly specify that a witness statement should set out a witness’ own recollection of facts and matters. By way of example, in PD 57AC para 4.1, a trial witness statement must be verified by a statement of truth which includes the words, “This witness statement sets out only my personal knowledge and recollection, in my own words.” However, the pitfalls of AI mean that accidental noncompliance with the CPR is entirely possible, as illustrated by the following cases:
- In Godwin v Godwin [2026] EWHC 923 (Ch), the Defendant and his former wife used ChatGPT in the preparation of their witness statements to assist with grammar, spelling and presentation. The Defendant’s witness statement described ChatGPT as a “digital assistant” (which the Judge thought was unconventional English and likely generated by AI), and claimed that ChatGPT had not added, removed or rearranged any of the words in the first draft of their witness statements which they had themselves prepared. The Judge considered that there was “no good reason” for either witness to have used ChatGPT, being sophisticated people who could have used other spelling or grammar checks like Word. In the absence of the initial drafts, the Judge therefore stated he could not be sure the witness statements were “in their own words, reinforcing my decision to approach the Defendant’s evidence cautiously”.
- In R v FGD [2026] EWCA Crim 918, it emerged mid-trial that the prosecution witness had used AI tools to prepare for cross-examination. Using AI, she had created a summary record of what she recalled of events on the night in question, followed by a series of questions in the form of a cross-examination of her account, together with suggested answers. In substance, this was witness coaching by an AI tool, and the defence applied to stay the proceedings for abuse. The Court of Appeal ultimately directed that there be a retrial. While there had undoubtedly been witness coaching, the witness’ evidence in chief had been given years before trial and therefore was not contaminated by the coaching content of AI. Further, because the AI “coach” to the witness was fully recorded and could be placed before the jury, it was therefore possible to see what had been said, or suggested, to the witness.
These cases highlight the importance of judicious use of AI tools. In line with existing common law principles, the courts in both cases were at pains to understand whether the evidence arose from the witnesses’ own words, or whether anything had been generated or changed by AI. It is worth emphasising that while the use of AI tools themselves is new, the legal principles which underpin these cases are not.
On a more practical point, if court users are determined to make use of AI when drafting their witness statements, it is important to maintain initial drafts of their statements and to keep a complete record in order to evidence the extent of their use of AI. Solicitors may also wish to warn witnesses, who will understandably be nervous about being cross-examined, not to use AI to coach themselves.
Possible changes to the rules of the court
However, it is clearly important to examine whether rules are needed to govern the use of AI for the preparation witness statements. A Working Group established by the Civil Justice Council (“CJC”) in January 2025 has produced an interim report focusing on AI that generates substantive content.
In relation to witness statements covered by PD57AC and trial witness statements, the Working Group’s interim report stated that it was difficult to see that the aims and objectives of PD57AC, or the rules in Part 32 of the CPR, could properly be met if AI is used to generate substantive text. It suggested requiring a declaration that AI had not been used to generate the content of a statement (even by way of rephrasing the witness’s evidence). As to non-trial witness statements, the interim report did not see any need for further rules as long as a legal representative bears professional responsibility for the preparation of the statement.
A mandatory declaration in similar terms is now a requirement for all affidavits and witness statements in the High Court of Ireland. On 29 July 2026, the “Practice Direction on the Responsible Use of Generative Artificial Intelligence in Court Documents” was published by the President of the High Court. It sets out prescriptive rules for the use of Generative AI (“GenAI”) in affidavits and witness statements:
- Particular caution must be exercised if GenAI is used to assist in the preparation of documents which form the evidence or potential evidence of a witness.
- Unless the Court otherwise directs, the maker of a witness statement or person swearing an affidavit must include a declaration stating that GenAI has not been used to generate the substantive content of the document (including by way or altering, embellishing, strengthening, diluting or rephrasing the evidence set out) and that the document reflects the witness’s personal knowledge, recollection and evidence.
- The use of GenAI for purposes unconnected to the generation of substantive content (such as formatting or spell-checking) does not prevent this declaration being made.
In June 2026, an update released by the Working Group stated that witness statements are the principal area of ongoing consideration, and they are giving careful consideration to whether existing mechanisms (statements of truth, disclosure obligations and cross examination) are sufficient to deal with concerns around the integrity of witness evidence. A draft report setting out its recommendations is expected soon, and it seems likely that the CJC will follow the High Court of Ireland and mandate declarations regarding the use of AI in the preparation of witness statements.
While AI usage that does not generate substantive content (like spell checking) is not an area of concern, the courts are clearly, and rightfully, wary of AI that substantively alters the content of witness statements which the court is asked to rely on. Given the recent cases surrounding the use of generative AI in witness evidence, clarity from the CJC will be welcomed by all court users.
© Anson Cheung, Ryan Fincham, 4 New Square Chambers, September 2026
This article is not intended as a substitute for legal advice. Advice about a given set of facts should always be taken.
