Civil Justice Council consults on new rules governing the use of AI in preparing court documents

The CJC refers to the judiciary’s revised guidance for judges on the use of AI (updated October 2025), which emphasises that legal representatives remain fully responsible for all material they place before the court.

Against this backdrop, the CJC is now consulting on whether current guidance remains adequate, or whether changes to the rules are required.

Key risks identified

The consultation paper highlights several forms of AI-related risk already visible in litigation. These include:

• “Hallucinations”: fabricated case citations or legal propositions generated by large language models.
• Biases and errors embedded within LLM training data.
• The growing sophistication of AI-generated text, images and video, raising concerns around authenticity of evidence.
• The potential for “white text” or hidden prompts to manipulate electronic review tools.

The paper also notes that jurisdictions across the world are grappling with similar issues with no clear consensus – some jurisdictions (such as the U.S. District Court for the Northern District of Texas) already require upfront declarations of AI use in filings, while others prefer a more principles-based approach.

Proposals relating to different categories of court documents

Statements of case

The CJC’s provisional view is that no new AI-specific rules are required, provided the statement of case bears the name of the legal representative taking professional responsibility for it. Requiring declarations of AI use could delay proceedings and, as AI becomes more embedded in legal workflows, may become impractical.

Skeleton arguments and advocacy documents

A similar approach is proposed. Although AI may increasingly assist with legal argument and citation-finding, the CJC considers that the existing professional duties of solicitors and barristers are sufficient. Provided the document clearly identifies the responsible lawyer, further rule-making is not currently justified.

Disclosure

Noting that AI-assisted disclosure is already well established, the CJC sees no present need for formal declarations of AI use in disclosure lists, noting that parties are generally cooperating effectively and issues have yet to reach the courts. The CJC does however accept that guidance may be required to deal with AI in disclosure in the future to cover issues such as the disclosure of prompts and the level of human oversight.

Witness statements

This is the area in which the CJC proposes the most intervention.

• Non trial witness statements: as with statements of case and skeleton arguments, existing professional responsibilities may be enough, with no additional rules required.
• Trial witness statements under PD57AC: given the Practice Direction’s emphasis on using a witness’s own words, the CJC proposes a mandatory declaration that AI has not been used to generate, edit, or rephrase substantive evidence.
• Trial witness statements under CPR Part 32 (outside PD57AC): a parallel declaration is proposed.
• Witness statements involving translation: the CJC asks whether rules should explicitly permit AI assisted translation, provided a human translator certifies accuracy, or whether broader use of publicly available machine translation should be allowed if identified.

Expert reports

Given recent cases in which experts inadvertently relied on inaccurate AI-generated content, the CJC proposes amending PD35 to require experts to identify and explain any substantive use of AI in their reports (other than for administrative tasks such as transcription).

Next steps

The consultation underscores the pace at which legal practice is changing. The CJC aims to maintain a balance: enabling the courts and practitioners to benefit from AI-driven efficiency while safeguarding the integrity of evidence and maintaining public confidence in the justice system.

The consultation remains open for responses until 14 April 2026.

Authored by Reuben Vandercruyssen, Andrew Holland and Alex Cumming.