In just one week there have been two different updates on the likely direction of travel for court procedures in view of AI; one from the Civil Justice Council’s Working Group and the other from the Technology and Construction Court. Amanda Savage KC and Helen Evans KC explain what’s going on and why .
The overall direction of travel
The CJC Working Group has been looking at how courts should respond to AI use across a range of work, including skeleton arguments, witness statements and expert reports, as well as use of AI in disclosure[1]. Its final report is not yet ready but in late June 2026, it published a brief Update.
The CJC’s June 2026 Update confirms that the “direction of travel” is:
- Maintaining the current position for AI use in professional legal drafting (i.e. it is not likely to be necessary for lawyers to sign declarations about whether AI has been used in producing documents such as pleadings or skeletons);
- Introducing “proportionate transparency” in relation to expert evidence; and
- Recognising the distinct and evolving challenges posed by litigants in person.
This is not entirely the same as the guidance on AI contained in the new TCC Guide, published on 1 July 2026, as explained below.
–
Legal drafting
It appears unlikely that the CJC Working Group will impose a requirement for Court documents (such as statements of case, skeletons) to contain a statement by the legal representative explaining whether and how AI has been used.
There are obvious advantages of this.[2] In particular, it is hard to see how any meaningful statement as to the use of AI can be given without the legal representative providing quite detailed information. This could be onerous and would be liable to intrude upon the legal representative’s “workings” or thought processes and infringe the party’s privilege.
On the other hand, it is clear that the use of AI in these documents is causing problems: there have been a significant number of cases recently heard in Court and in the Tribunals dealing with actual or suspected “hallucinations”[3]. This suggests that, contrary to the optimism in the CJC’s Interim Report and Consultation, the addition of the legal representative’s name on the document – which carries with it (or should carry with it) an acceptance by the legal representative of his or her professional responsibilities and compliance with professional obligations – is not yet providing sufficient.
Perhaps for this reason, the TCC Guide takes a more nuanced position. It says that “whether and to what extent legal representatives should disclose their use of AI will depend on the context. However, any doubt as to whether the fact that AI has been used should be mentioned must be resolved in favour of disclosing such use” (p. 9)[4]. Other jurisdictions have also made indications favouring transparency: see for instance the decision of the Irish Court of Appeal in Guerin v O’Doherty [2025] IECA 48 (at para. 74).
Witness statements
The CJC’s June Update also states that “the group is actively exploring whether a proportionate and workable approach can be developed for witness statements, including whether any form of disclosure would enhance, rather than complicate, the administration of justice”.
The use of AI for trial witness statements strikes us as problematic. The purpose of a trial witness statement is to give the evidence that the witness would give if giving oral evidence in chief, and it should be in their “own words”. The whole point of Practice Direction 57AC when it was introduced in 2021, was to move away from a mechanical recitation and analysis of documents and/or an “over-lawyering” of statements. It is hard to see how the use of AI to generate trial witness statements could be in keeping with that philosophy and the requirements of PD57AC- which are re-emphasised in the recent TCC Guide (p. 89).
The CJC seem to recognise this and propose there should be a rule requiring a declaration that AI has not been used for the purpose of drafting such witness statements.
The position seems to be different for non-trial witness statements where it is proposed by the CJC that the position should be in line with that for statements of case and skeleton arguments.
Experts
Experts already occupy a different position to legal representatives in their dealings with the court. For instance, they are expected to be clear about what data they have relied on, and what they have and have not reviewed (see e.g. CPR PD35 para 3.2-3.3). The TCC Guide provides that any calculations produced by software such as Excel should be disclosed, setting a precedent for disclosure of some tools (p. 101). So should AI use be any different?
The CJC’s reference to “proportionate transparency” in its June 2026 Update is intriguing. It suggests that the CJC will impose some type of formal, specific requirement that experts disclose their AI use. Whether that is done by amending the experts’ statement or producing guidance requiring a specific explanation within the body of the report (or by some other route) remains to be seen.
One thing that is clear is that cross-examination into the process by which expert reports have been produced is likely to increase- as has been the experience since the introduction of CPR PD 57AC governing the process for producing witness statements.
–
© Amanda Savage KC and Helen Evans KC, 4 New Square Chambers, July 2026
This article is not intended as a substitute for legal advice. Advice about a given set of facts should always be taken.
[1] For a helpful summary see the article by Isabel Barter and John Williams: AI and the Courts: The Civil Justice Council’s Consultation on Preparing Court Documents – 4 New Square Chambers
[2] Some of which are flagged in the Interim Report and Consultation.
[3] For instance: R (Ayinde) v LB Haringey, Al-Haroun v Qatar National Bank [2025] EWHC 1383, UK and R (on the application of Munir) v Secretary of State for the Home Department [2026] UKUT 81 and Cork and another v Mark Smith [2026] EWHC 1199 (Ch).
[4] It then goes on to remind legal representatives of the need to protect confidentiality in any data in respect of which AI is used and to check the accuracy of any research of case citations generated by AI>
