Light-Touch Regulation of AI: What the CJC’s Final Report Means for Practitioners

AI is increasingly being used to prepare court documents, but how far should its use be regulated? 

On 29 October 2026, the Civil Justice Council (CJC) published its final report on the use of AI in preparing court documents, with recommendations designed to protect the integrity of civil proceedings without imposing unnecessary procedural requirements.
The report follows an interim report and consultation, drawing on input from a Working Group chaired by Sir Colin Birss and bringing together expertise from across the judiciary, the CJC, the Law Society and the Bar Council. It makes fifteen recommendations and offers an important indication of the direction of travel for AI in civil litigation.

For lawyers as a whole, the report proposes a light touch approach, with intervention only where there is a real need. The report recommends no new AI declarations for statements of case, skeleton arguments or other advocacy (recommendation 4), no disclosure reform for now (recommendation 5), no duty to name AI tools to the court (recommendation 14) and no permission regime (recommendation 15). With the underlying theme being that responsibility stays where it already sits: with the named lawyer who takes responsibility for the document. On translation, it asks that certification requirements be made explicit for AI translations, in addition to requiring human verification of AI translations and providing the original text (recommendation 8).

The report’s core principle is that rules around AI should be functional and risk-based. The rules should turn on what an AI tool does, and not what type of tool it is. The question is whether it can “generate” or “influence” the substantive content of what goes before the court: facts, evidence, opinion, reasoning or argument contained in the document. A spell-checker or similar tool would fall outside of that, whilst for AI transcription the risk depends on whether the use generates or affects content - whilst it may be administrative in nature, it can still fabricate content and therefore fall into the category of generating or influencing substantive content that is put before the court.  Put another way, function and risk is what is important, rather than the categorisation of a tool. 

The report furthermore covers other key proposals as follows:

• Codification of approach in new Practice Direction (recommendation 3): The Working Group recommends that CPRC draws together its recommended functional, risk-based approach and the case law on AI misuse and sanctions, such as hallucinated authorities, into a plain-language Practice Direction. It should also give guidance on specific contexts, including witness statements and disclosure (subject to the Disclosure Review Working Group’s work on PD 57AD).
• Witness statements (recommendations 6 and 7): The Working Group also makes a recommendation for witnesses of fact to make a declaration that AI has not generated or influenced the substance of their statement, so that it remains in their own words. The report likens AI embellishment to witness coaching. However, statements prepared and signed by legal representatives (except where, they themselves are a witness of fact) will not require a declaration of AI use, given that responsible use would fall within their existing professional duties.
• Expert reports (recommendation 9): The Working Group also makes a recommendation for experts to make a declaration where AI has generated or influenced their report. The declaration must include sufficient information (and, where necessary, the tool used) to enable the opposing party and the court to scrutinise the methodology and appropriateness of its use. This is a distinct approach from legal representatives, as experts’ methods must be testable.

Whilst litigants in person were outside the Terms of Reference, the report calls for the CJC to address AI use by litigants in person urgently (recommendation 2), given the absence of guidance.

The CJC recommendations are timely given that publication coincided with the update to Matthew Lee’s UK AI Hallucinations Cases Tracker breaking through the 100-case threshold.  Whilst the report’s overall recommendations are light-touch, it is still important for our general understanding of AI use in the civil proceedings world, moving forward. Whilst nothing changes yet, the direction is clear: use AI if it helps, verify everything, and be ready to stand behind your name on the document.