Witness Statements, AI Transparency and What Courts Are Starting to Demand

The Direction of Travel Is Clear

The Civil Justice Council is actively exploring what it calls proportionate transparency around AI use in witness statements, and the Court of Appeal has now signalled that criminal proceedings deserve the same scrutiny. Taken together, these are not isolated procedural curiosities. They represent a coordinated judicial reckoning with the reality that AI tools are already inside the litigation process, whether courts formally acknowledge it or not.

For legal professionals, the message is straightforward: the question is no longer whether AI was involved in preparing a document, but whether that involvement was disclosed, understood and appropriate. Courts are beginning to build the frameworks that will make undisclosed AI use a procedural risk rather than a quiet efficiency gain.

Why Witness Statements Are the Flashpoint

Witness statements occupy a specific and sensitive position in litigation. They are meant to represent the unmediated recollection of an individual. When a generative AI tool is used to structure, refine or, in the worst cases, embellish that recollection, the integrity of the evidence itself is potentially compromised. The concern is not hypothetical. Lawyers and litigants under time pressure have obvious incentives to use AI to produce polished, coherent statements quickly. The risk is that the polish obscures the seams between genuine recollection and AI-generated narrative.

The proportionate transparency standard being explored by the CJC suggests that courts are not looking to ban AI assistance outright. That would be both unenforceable and probably unnecessary. What they want is a disclosure regime that lets the court assess the degree of AI involvement and weigh the evidence accordingly. The shape of that regime, including what must be declared, by whom and at what stage, remains to be defined.

The Governance Gap This Exposes

For in-house legal teams and law firms, these developments expose a governance gap that many organisations have not yet closed. Most firms now have some form of AI use policy, but those policies tend to focus on generative AI in drafting advice, deal documents or correspondence. The use of AI in litigation support, including witness preparation, chronology building and document review, is often governed less formally or not at all.

If a witness statement is later challenged on the basis that AI was used in its preparation without disclosure, the question will quickly turn to what oversight existed, who knew and what records were kept. Firms without clear processes will find that question very uncomfortable indeed.

What Responsible AI Use in Litigation Actually Looks Like

Responsible use does not mean avoiding AI. It means deploying it with a clear audit trail, keeping the human author genuinely in control of the substance, and being prepared to disclose the role AI played if the court or the opposing party asks. Practically, this requires a few things that good practice should already demand.

First, any AI tool used in the preparation of litigation documents should be logged at the matter level. Second, the supervising lawyer must be able to attest honestly that the factual content originates with the witness, not with a language model. Third, firms should begin updating their file management processes now so that AI involvement is captured as a matter of routine rather than reconstructed under pressure.

The principle here is not far from the one that governs AI use in contract drafting. When Adira drafts or reviews a contract, the output is always anchored to what the client actually agreed, the jurisdiction that governs the deal and the company's own established positions. The human lawyer or commercial team retains authority over the substance. The AI handles the structural and linguistic work. That division of labour is clear, auditable and defensible. Litigation teams need an equivalent clarity.

What to Do Before the Rules Arrive

The CJC and the Court of Appeal are still in the investigative and consultative phase. Formal rules, whether in the form of practice directions or judicial guidance, will take time to crystallise. But waiting for final rules before establishing internal governance is the wrong approach. The reputational and professional conduct risks of getting this wrong are present now, and the courts are already alert to the issue.

In-house teams should begin by auditing how AI tools are currently being used across their litigation matters. Law firms should review their AI use policies to ensure litigation support is explicitly covered. Both should consider what a disclosure statement would look like if a court asked today, and whether they could produce one confidently.

Transparency in AI use is becoming a baseline expectation across the legal system, not just in courts but in contracts, regulatory filings and commercial negotiations. Organisations that build transparent, well-governed AI workflows now will be better placed when formal obligations arrive, and better protected if a challenge arises in the meantime.