Law Society response to CJC AI consultation: guidance first, rules later

The CJC consultation asked whether new procedural rules are needed to regulate how legal representatives use AI when preparing statements of case, skeleton arguments, witness statements, expert reports, translations and disclosure documents.

The consultation comes against a developing background of judicial and professional guidance. Courts and regulators have increasingly stressed that lawyers remain responsible for material placed before the court, even where technology has been used to assist with research, drafting or review. Recent cases involving fabricated citations and inaccurate AI-generated legal material have also increased attention on verification, supervision and professional responsibility.

The Law Society's response accepts that existing professional duties remain important. Solicitors already owe duties to the court, to clients and to the proper administration of justice. But the Law Society does not see those duties as enough on their own. Its view is that professional responsibility and transparency are separate points. A lawyer's name on a document may confirm that there is an accountable person, but it does not tell the court or the other parties whether AI was used, how it was used, or whether further checks may be needed.

What the Law Society recommends

The response first calls for SRA guidance. The Law Society says the SRA should “issue guidance soon” on how solicitors' professional duties apply when AI is used in preparing court documents. It says this should include expectations around “baseline AI competence”, including an understanding of AI's “capabilities, limitations and appropriate verification techniques”.

The Law Society also suggests that the SRA may need to review the Code of Conduct. It is concerned that individual solicitors should not be unfairly penalised where they could not reasonably have known AI had been used, or where they were working without adequate support, training or controls. Its recommendation is that “there should not be penalties” where a solicitor could not reasonably have been aware of AI use.

This reflects two practical points. First, AI may be used in different ways at different stages of preparing a document, from research and drafting support to case management systems and third-party workflows. Secondly, in larger legal teams, the solicitor taking responsibility for a document may not always know precisely how AI has been used by paralegals, trainees, junior lawyers or others involved in preparing the material.

The response also calls for HMCTS guidance. The Law Society proposes “short, principles-based guidance” for all court users. That would include litigants in person, experts, translators and others involved in proceedings. It suggests that a wide range of principles may be relevant, including “transparency, explainability, equality, confidentiality, human review, accuracy, security and standardisation”.

That guidance could include basic recommendations on declarations of AI use, checking AI outputs, avoiding public chatbots for confidential material and identifying circumstances where AI should not be used. The Law Society expressly links confidentiality to a possible “direction not to use publicly available chatbots”, quoting the judicial guidance warning not to enter information into a public AI chatbot unless it is already in the public domain.

The Law Society also recommends a wider MoJ review. It says AI use in court documents cannot be looked at in isolation and calls for a review covering AI use by court administration, the judiciary, parties, litigants in person and relevant third parties. Its preliminary view is that future CPR amendments should include “new transparency obligations”, but it reserves its position until the effect of SRA and HMCTS guidance is clearer.

Transparency and document-specific proposals

Transparency is the strongest theme in the response.

The Law Society disagrees with the CJC's proposal that statements of case and advocacy documents need no further AI-specific rules if they identify the legal representative taking responsibility. It says this would not provide the “degree of transparency necessary” for parties to satisfy themselves of fair process. Its preferred approach is a “proportionate transparency requirement” focused on “material uses of AI” that are reasonably known to the legal representative and capable of being verified.

It does not propose disclosure of every AI touchpoint. Straightforward spelling, grammar and basic formatting should generally fall outside any disclosure requirement. But the response says AI use should be disclosed where it generates text setting out evidence or factual assertions, produces graphics, images or videos, supports legal research, generates legal argument, or is used for transcription, translation or accessibility functions where there are risks of inaccuracy, bias or unfairness. 

The response does not, however, explain how a distinction should be drawn between basic and more complex uses, or how that distinction may change over time as AI tools become more capable of performing complex tasks.

The Law Society takes a stricter approach to witness statements. Its position is that all witness statements should include a declaration of non-AI use, and that AI should not be used for the purpose of “altering, embellishing, strengthening, diluting or rephrasing” a witness's evidence. 

On disclosure, the Law Society says a “statement of AI use is preferred”. It recognises that technology-assisted review is already familiar in large-scale disclosure, but notes that “more advanced and agentic uses of AI” may justify a basic transparency requirement, potentially in the Disclosure Review Document or disclosure statement.

For translations, it does not call for a specific rule governing AI use by human translators, but says “a statement of accuracy alone will not suffice” and recommends a statement of AI use alongside the translator's statement of accuracy.

For experts, the Law Society supports amending statements of truth so that expert reports identify and explain substantive AI use, other than purely administrative uses. It also says court guidance should provide “red lines” on when an expert can use AI, with case examples.

Analysis

The response is a useful marker of opinion across the solicitors' profession. The Law Society represents solicitors in England and Wales as a whole, and its position reflects different levels of AI adoption, enthusiasm and sophistication across the market. Some firms are already using AI in structured workflows; others have limited exposure and may encounter AI mainly through litigants in person, opponents, experts or third parties.

That breadth helps explain the cautious tone. The Law Society does not suggest that AI should be excluded from legal practice. Its concern is that AI use should be supported by competence, controls, checking and proportionate disclosure. It also recognises the risk of a two-track environment, where some parties can use sophisticated tools safely while others rely on public systems or have little understanding of the risks.

The response goes further than the CJC consultation in some respects. The CJC's provisional approach was, in several areas, to rely on existing professional responsibility and identification of the lawyer taking responsibility for the document. The Law Society instead favours additional guidance and, in due course, transparency obligations for material AI use. Some will see that as a sensible response to a fast-moving technology. Others may say parts of the proposal go too far, particularly where English litigation already contains structured transparency mechanisms, such as the Disclosure Review Document in disclosure exercises.

The “proportionate transparency” proposal will divide opinion. A declaration of material AI use may help the court and other parties identify whether further scrutiny is needed. But it also raises questions about how far procedural transparency should reach into the lawyer's drafting and preparatory process, and how it interacts with confidentiality and privilege. The Law Society's formulation – material uses of AI that are reasonably known to the legal representative and capable of being verified – is a pragmatic attempt to deal with larger teams and embedded tools. It does not remove the continuing need for lawyers to supervise and manage AI use appropriately.

The response leaves some difficult questions for later. It does not provide a detailed privilege analysis, beyond treating public chatbots as a confidentiality risk. Nor does it fully define what counts as “material” AI use, what a lawyer can reasonably be expected to know, or how much detail a declaration should contain.

The Law Society treats those questions as issues for further guidance and review, rather than matters to be resolved now. That reflects a wider pattern among regulated professionals and businesses – seen in areas such as financial services – of asking regulators for more specific AI guidance, and resisting the idea that existing frameworks are sufficient on their own.

AI use in civil litigation is unlikely to be banned, but it is also unlikely to be left entirely to informal practice. The emerging baseline is responsible use, human accountability and a clearer record of when AI has contributed to material placed before the court. 

Authored by Reuben Vandercruyssen, Andrew Holland, and Antonia Croke.