Writing in NLJ this week, Karl Hopkins, an independent researcher and public-service practitioner, identifies a ‘source-integrity gap’: witnesses may honestly write their own statements after absorbing machine-generated explanations or suggested answers.
‘Authorship can be verified, while the origin of the account cannot,’ he warns.
Hopkins argues that the Civil Justice Council’s proposed declaration on AI-generated witness statements must be accompanied by a positive duty to ask about earlier substantive use.
Solicitors should make written enquiries before the first conference, repeat them before signature and record the answers. Relevant prompts, outputs and chat histories should be preserved proportionately, subject to privilege.
He distinguishes administrative assistance from AI supplying missing facts, reconciling disputed events or rehearsing answers. His proposed safeguard targets that substantive interference, protecting recollection before the drafting process obscures its origins.
