Significant advancements in the text generation and processing capabilities of AI tools have raised questions about how AI could streamline the production of witness statements. Civil Procedure Rule 32.8 and Practise Directions (PD) 32 and 57AC (the latter focusing on trial witness statements in the Business and Property Courts) require witness statements to adhere to specific requirements of content and substance, which need to be carefully considered should AI tools be implemented in the drafting process.

Requirements of content and substance

Pursuant to PD 32.18.1 and PD 57AC 3.3, a witness statement ought to be articulated in the witness' 'own words' and 'own language', requiring full understanding and approval of the witness. Solicitors should avoid re-forming the witness' recollection/understanding of events in their own language to ensure the witness retains ownership of the statement. Consequently, it is unlikely a witness could simply input prompts into an AI tool and pass off the output as their 'own words'. The very nature of generative AI means that it learns from countless sources of written material, combining and condensing into a style definitively not that of any one user.

Further, PD 57AC (i) requires transparency of the drafting process and (ii) mandates that statements reflect a witness' own knowledge. Automating drafting through AI means an intermediary is inserted between witness and statement, severing the direct link between original testimony and final draft. In so doing, it calls into question if the draft is truly the full and entire recollection of the witness. Moreover, PD 32.18.1(5) requires disclosure of how a statement was prepared (regardless of the court) - evidently drawing attention to the involvement of AI, opening the door to a suggestion that the statement is not comprised solely of the witness' own recollection.

Could AI tools be used to draft witness statements?

On the other hand, appropriate AI tools could be utilised to increase the efficiency and accuracy of drafting statements:

• Transcription and editing – AI could be used to transcribe interviews conducted by lawyers with the witness before editing the text into a draft witness statement, keeping the text as close to verbatim as possible in order to retain the witness' own words. The end result would still be subject to the usual review by the witness, but this approach could well expedite the process and cut costs.

• Automated document preparation – AI tools could be used to collate and chronologise the materials that ought to be put in front of the witness during the drafting process to reduce lawyers' preparation time. However, this use of AI would need to be subject to the usual checks and balances due to risks of both bias and error.

• Unbiased interview technique – poor interview practice, for example asking leading questions, can lead to the interviewer's biases and assumptions altering the responses of an interviewee, possibly even contributing to the creation of 'false memories'. AI tools can be used to conduct a 'cognitive interview': a structured approach to interviewing that adopts a neutral, fact-finding approach.

Conclusion

PD 57AC was introduced to streamline the witness statement process, reducing both cost and legal overengineering. This suggests that the judiciary are fundamentally open to the implementation of mechanisms (such as AI tools) which can increase the efficiency of evidence preparation. Whilst the requirements for statements to be in a witness' 'own words' and reflective of personal knowledge are difficult to reconcile with the use of AI when drafting, there is potential for using AI tools in transcription, document organisation, and structuring of interviews – all of which can go some way towards cutting costs and improving efficiency.

Should you require any further information on the issues covered in this article, please contact one of our Disputes and Investigations team.