Authors
Recent Court developments
Godwin v. Godwin
The use of AI to draft witness statements was recently heavily scrutinised in Godwin v. Godwin [2026] EWHC (Ch) 923 (April 2026), where the High Court, Chancery Division, considered witness statements that had been prepared with a “digital assistant”, and that transpired to be ChatGPT. The defendant and his former wife said they had initially drafted the statements in Word themselves before turning to ChatGPT for grammar, spelling, and presentation only. The Court nevertheless questioned whether the statements were in the witnesses’ own words, highlighting the absence of the original drafts and the difficulty of testing what the tool had changed [44]-[46]. The Court also scrutinised the description of the tool, concluding that it was not conventional in spoken English to refer to an AI tool as “a digital assistant” and that technically sophisticated witnesses had no compelling reason to use ChatGPT when standard spell-checking tools existed. Therefore, in the absence of a satisfactory explanation and noting that the Court could not be sure that the witness statements were the witnesses’ own words (contrary to the Civil Procedure Rules), the Court concluded that the evidence had to be approached cautiously. Once AI has intervened, the Court may not be able to separate cosmetic assistance from substantive influence after the event.
PD57AC requires witness statements to be in the witness’s own words and requires the witness to confirm that they understand their duty to the court.
R v. FGD
Use of AI by witnesses in the form of witness coaching was recently considered by the Court of Appeal, Criminal Division, in R v. FGD [2026] EWCA (Crim) 918 (June 2026). During cross-examination, the complainant produced screenshots to address a challenge about their whereabouts on the night in question. That led to an order for the complainant’s phone to be examined, which revealed AI-generated mock cross-examination-style questions with suggested answers. The Court of Appeal confirmed that this constituted impermissible coaching under R v. Momodou [2005] EWCA (Crim) 177, holding that AI coaching falls squarely within the prohibition on witness training, which is “forbidden in all jurisdictions” [23]. AI-generated chronologies, suggested answers, and mock cross-examination risk changing how an account is remembered or presented in civil proceedings, too.
R v. Momodou draws a practical distinction between witness familiarisation and witness coaching. Familiarisation is helping a witness understand the process and courtroom procedure. That is permitted, provided it does not influence the substance of the evidence. Coaching, on the other hand, including rehearsing answers or shaping testimony, is prohibited.
Ultimately, FGD demonstrates that these forensic implications warrant emphasis. Using AI creates a risk that the Court might consider excluding evidence and might also expose a user to an unwelcome, wider-ranging, and potentially more intrusive investigation and examination for disclosure than would otherwise have been required.
The CJC proposals
On 30 June 2026, the Civil Justice Council published an update on its consultation regarding artificial intelligence in court documents. The review found a strong consensus that existing professional responsibility frameworks are sufficient to govern appropriate AI use in preparing pleadings, skeleton arguments, and advocacy documents. The June 2026 working update followed the Interim Report.
Under the Interim Report, the core proposals under consultation were:
- Trial witness statements: A mandatory declaration that AI has not been used to “generate the content” of the statement. This includes altering, embellishing, strengthening, diluting, or rephrasing the witness’s evidence. The declaration would supplement PD57AC, which already requires a certificate of compliance from the legal representative, and would make the witness’s assurance an express procedural requirement. Notably, non-text-generating AI tools (e.g., grammar checking, transcription) would be exempt.
- Expert evidence: The consultation asks whether the specific form of statement of truth used by experts to verify their reports should be amended to specifically refer to AI. It also asks whether there should be a requirement that the expert explain, in the body of the report, what use of AI has been made, identifying the AI tools that have been used.
- Machine translation: The consultation asks whether publicly available machine translation services should be permitted for foreign-language witness statements, provided the tool is identified and opposing parties can verify the translation. An alternative would be to permit such use to legal representatives but only with a requirement that the legal representative identifies what tool is used.
- Human translation: The consultation asks whether there should be a rule which allows AI to be used by human translators. If the translator is prepared to sign a statement of accuracy, taking responsibility for the AI, the consultation asks whether there is any need to enquire further.
- Scope of AI: The consultation asks whether the term “AI” is sufficiently clear or whether it needs to be defined further.
Conclusion
These developments carry a clear message: Practitioners must take proactive steps to manage the risks of AI in witness evidence. At the outset of every instruction, legal representatives should establish explicit protocols setting out which AI tools witnesses and experts may use and in what circumstances. Witnesses should be warned that undisclosed or improper use of AI may expose the case to intrusive disclosure orders, undermine the weight given to an expert’s evidence, or result in that evidence being excluded altogether. As the regulatory framework continues to develop, firms should audit their current practices now to ensure that they are ready for the mandatory declarations and enhanced scrutiny that lie ahead.
Client Alert 2026-177