Geoffrey Vos, the outgoing Master of the Rolls, is using his final weeks in office to spark a debate on artificial intelligence in the English civil justice system. He routinely relies on Microsoft Copilot to locate distant documents and condense large texts, and he recently toured a Shenzhen court where judges already depend heavily on AI assistance for civil disputes.
Current UK guidance restricts judges to using AI for purely administrative purposes, yet Vos says he employs Copilot daily for research, navigation of extensive material, and verification tasks. He notes that the tool can retrieve a decades-old document relevant to a meeting within seconds, a capability that saves him considerable time given his eleven-meeting daily schedule.
The rapid spread of AI in legal practice has produced notable risks. Multiple lawyers have been caught citing case-law that was fabricated by AI, and a UK practitioner recently lost their licence after misusing the technology. At the same time, litigants are using chatbots to draft claims, contributing to a ten-percent annual increase in county-court filings and a flood of potentially frivolous cases.
Vos warns that AI-generated evidence, such as deepfake images, video, or audio, adds a new layer to an age-old problem of forged material. He draws a parallel to historical forgeries of wills and recordings, noting that courts have always required judges to assess authenticity and that perjury charges will still apply to fabricated digital evidence.
Despite these challenges, Vos believes that certain categories of civil disputes,personal injury, low-value professional negligence and similar claims,could eventually be resolved entirely by AI. He argues that delivering justice in a way that matches contemporary cultural expectations will require broad consensus, possibly through legislation, to ensure that algorithmic decisions are accepted by the public.
The judge stresses that the core issue is not the opacity of AI reasoning but the plausibility of its outcomes. While the internal logic of large language models may remain hidden, their generated reasoning can be evaluated for credibility. Nonetheless, he acknowledges that human judges fulfill a deep-seated psychological need for peer judgment, suggesting that AI will complement rather than replace human adjudication in the foreseeable future.