AI in the Witness Box: What the Courts' New Scrutiny Means for How Organisations Use Legal AI

The Courts Are Paying Attention

The Civil Justice Council is actively exploring what it calls "proportionate transparency" around the use of AI in preparing witness statements, and the Court of Appeal has now recommended a parallel investigation for criminal proceedings. Taken together, these two developments mark a meaningful shift in judicial attitude: AI assistance in legal preparation is no longer an invisible background activity that practitioners can quietly rely upon. It is becoming something courts expect to see disclosed, scrutinised and governed.

For litigation teams this is immediately significant. But the implications run considerably wider than the courtroom.

Why Witness Statements Are a Useful Lens

Witness statements occupy a peculiar position in legal process. They are meant to record what a person actually recalls, in their own words, with their own emphasis. When AI assists in drafting or refining that account, a fundamental question arises: whose voice is this, and how faithfully does it represent the witness's genuine recollection rather than a statistically optimised version of it?

The courts are right to ask that question. But the same question has a quieter analogue in transactional work. When an AI system drafts a contract clause, negotiates a position, or summarises obligations, it is also producing output that will be attributed to and relied upon by a human party. The standard of care attached to that output, and the transparency owed to counterparties and decision-makers, deserves the same rigour.

What This Signals for In-House Teams and Law Firms

The emerging judicial consensus points toward a world in which AI involvement in legal documents will need to be disclosed and, where challenged, defended. In-house general counsel and their external advisers should read this as an early signal to do three things now rather than under pressure later.

First, build a clear internal record of where and how AI tools are used in document preparation. That does not mean abandoning efficiency; it means creating audit trails that demonstrate human review and ultimate human accountability for every significant output. If a court or regulator ever asks, the answer should be ready.

Second, distinguish between AI that assists human thinking and AI that substitutes for it. A tool that suggests clause language which a qualified lawyer then critically reviews and amends sits in a very different risk category from one whose output is accepted wholesale. The courts' concern about witness statements is essentially a concern about substitution without scrutiny. The same concern applies to contracts.

Third, revisit your matter management and CLM workflows to ensure that the human sign-off step is genuinely meaningful. Clicking approve on an AI-generated document without substantive review is likely to attract exactly the kind of judicial and regulatory disapproval that is now being articulated for witness statements.

Where Adira's Approach Fits

Adira is built on a principle that matters here: AI should work from your side of the table, in your organisation's voice, and with a working knowledge of the law of the jurisdiction it operates in. Those are not marketing points. They are design choices that directly address the transparency problem the courts are raising.

When a clause is drafted in your organisation's established contractual language, reviewed against the governing law of the contract, and surfaced to a qualified reviewer with clear reasoning attached, the chain of accountability is visible. There is no mysterious black-box output that a counterparty or a court would struggle to interrogate. The human professional can explain every material choice because the tool has explained its reasoning as it goes.

This is precisely the kind of proportionate transparency the Civil Justice Council is reaching for, even if it is articulating it in the specific context of witness evidence.

The Broader Principle: Accountability Does Not Disappear Because a Machine Helped

The deeper point in the courts' current thinking is a simple one: using AI does not transfer accountability away from the professional or the organisation that deploys it. A witness cannot disclaim responsibility for their statement because an AI helped phrase it. A lawyer cannot disclaim responsibility for contract terms because a system generated the first draft.

This is not a reason to avoid AI in legal work. It is a reason to use it thoughtfully, with governance structures that make the human role clear and auditable. Organisations that treat AI as a productivity shortcut without accountability infrastructure are accumulating a risk they may not yet be pricing correctly.

The courts are beginning to price it for them. Getting ahead of that curve is straightforwardly good legal risk management.