When Witnesses Meet AI: What the Courts' New Scrutiny Means for Legal Teams

The Courts Are Paying Attention

The Civil Justice Council and the Court of Appeal are both moving toward what regulators are calling "proportionate transparency" around the use of AI in preparing witness statements. That phrase deserves unpacking. It does not mean a blanket prohibition. It means courts want to know when AI has been involved, to what degree, and whether the human witness still stands fully behind the content as their own evidence.

For anyone watching the development of legal AI governance in England and Wales, this is a significant moment. Witness evidence sits at the heart of adversarial proceedings. If AI tools can quietly reshape the language, structure or even the substance of what a witness says, the integrity of that evidence becomes uncertain. Courts are right to take this seriously, and legal teams should take the signal seriously too.

The Distinction Courts Are Trying to Draw

There is a meaningful difference between using AI to check grammar in a statement and using it to reconstruct events or suggest facts. There is also a difference between a tool that passively tidies prose and one that actively prompts a witness with leading summaries drawn from documents they may not have read carefully themselves.

The difficulty is that most AI tools in circulation today sit somewhere in the middle of that spectrum, and users do not always appreciate how much the output has drifted from the original input. A witness who dictates a rough account and then accepts an AI-polished version wholesale may genuinely believe the statement is their own, even if the framing, emphasis and vocabulary have shifted considerably.

Courts will increasingly expect solicitors to advise clients on this risk, to ask probing questions about how a statement was prepared, and to disclose AI involvement where it is material. Firms that have not yet built that advisory habit into their practice will need to do so.

What This Means for In-House Teams

In-house lawyers often sit at the beginning of the evidence chain. When litigation arises, it is frequently the in-house team that collects initial accounts from employees, reviews correspondence and begins shaping a narrative before external counsel is engaged. If AI tools are embedded in that workflow, and they increasingly are, the question of disclosure starts there.

This is not a reason to avoid AI. It is a reason to be deliberate and documented about its use. Teams should be able to say, clearly and honestly, which parts of any draft were AI-assisted, which were reviewed by a qualified lawyer, and where the witness themselves reviewed and confirmed the substance. That kind of audit trail is not just good governance; it will become a practical necessity in contested proceedings.

In-house teams should also be alert to the fact that this scrutiny will likely extend beyond formal witness statements over time. Chronologies, disclosure narratives and even instructions to counsel can all be AI-assisted, and courts may eventually take a broader interest in the provenance of litigation documents generally.

Where Adira's Approach Fits

Adira is built around a core principle: AI should work from your side of the table, in your voice, with full awareness of the legal context it is operating in. That principle is directly relevant here.

When Adira assists with document preparation, it does so with the user's instructions and jurisdiction in mind, not by generating plausible-sounding content that a human then rubber-stamps. The distinction matters enormously in a regulatory environment that is beginning to ask hard questions about AI provenance.

For contract work specifically, the parallel to the witness statement debate is the question of whether a contract genuinely reflects a party's negotiated position or whether an AI has quietly introduced language that neither side properly scrutinised. Adira's drafting model, which reads agreements from your perspective and flags issues against your own playbook, is designed precisely to keep the human lawyer in control of the substance, not just the final signature.

The Broader Governance Lesson

What the Civil Justice Council and Court of Appeal are doing is the beginning of a normalisation process. Over the next few years, disclosure of AI use in legal proceedings will almost certainly become routine, possibly mandatory in specific circumstances. Firms and in-house teams that treat that disclosure as a burden are thinking about it the wrong way.

Transparency about AI use is, in fact, a professional advantage. It signals that your team understands its tools, has appropriate oversight in place, and can account for how its work product was produced. The lawyers who will struggle are those who have adopted AI quietly, without governance frameworks, and who cannot now explain what their tools actually did.

The courts are asking a reasonable question. The answer should never be: we are not sure.