© Getty images
Generative AI can shape a witness’s account before a solicitor ever sees it: Karl Hopkins argues the profession’s current response does not go far enough
- Generative AI can compromise a witness’s evidence before drafting begins by shaping recollection or supplying explanations, creating a ‘source-integrity gap’ that authorship checks cannot detect.
- Lawyers should ask witnesses specifically about earlier substantive AI use, record and preserve relevant interactions, and prohibit AI from supplying or reconciling facts rather than merely assisting with administrative tasks.
In Godwin v Godwin [2026] EWHC 923 (Ch), two witnesses said they had used ChatGPT only to correct grammar and spelling, without adding, removing or rearranging their own words. Neither could produce the original drafts. Judge Klein was not satisfied that the resulting statements remained in the witnesses’ own words, and approached their evidence with corresponding caution.
Both parties in Godwin were litigants-in-person (LiPs), but only one of the two ChatGPT-assisted witnesses was a party; the other was the defendant’s former wife.
