England’s top civil judge urges courts to consider AI-decided cases—with consent

Some civil disputes should be considered for AI decision-making, not just AI-assisted paperwork, Geoffrey Vos argues in a WIRED interview published October 28. England’s most senior civil judge, who is days from retirement, says that shift would require consent in individual cases or authorization through legislation—not simply a decision by judges to adopt the technology.

From finding documents to deciding disputes

Vos, the outgoing Master of the Rolls, spoke with WIRED’s Joel Khalili at the Royal Courts of Justice in London in October. His position is a call for public debate about how courts should resolve disputes, rather than an announcement that England’s courts are handing judgments over to AI.

His own use remains on the assistance side of that divide. Vos says he uses Microsoft Copilot every day to find documents, navigate large amounts of material and check information. He does not use it to write judgments, and says he is more cautious about using it to read papers for his cases.

Current guidance for judges in England advises limiting AI use to administrative tasks, WIRED reports. Vos’s argument goes further: he believes AI could eventually deliver decisions in some civil cases more quickly and at lower cost. A recent visit to a court in Shenzhen, where judges used substantial AI assistance in deciding civil disputes, helped shape his enthusiasm.

It will only be possible to ask people to accept algorithmic decision-making if they consent to the process in a particular case, or if society decides through legislation.

Geoffrey Vos, speaking to WIRED

Which cases—and whose decision?

Vos does not propose replacing every human judge. He draws a distinction between disputes where people may prefer a quick resolution and cases where they expect a person to exercise judgment. The examples he discusses put different kinds of civil cases on either side of that line:

• Potential AI decisions: personal injury claims and low-value professional negligence disputes are among the cases Vos believes could be adjudicated by AI.
• Human judgment: custody and family cases, along with complex commercial disputes, are areas where he says people still expect a human judge.

Even in complex commercial litigation, he sees room for AI to collate information and make human decision-making easier. But he says judges are not yet in a position to set the boundary themselves. That choice belongs to society, politicians and legislation.

Acceptance is a separate problem from speed. Vos acknowledges that people may find an adverse machine-driven decision harder to accept than one delivered by a human. Asked about understanding how a model reaches its answer, he puts greater weight on outcomes and the plausibility of the reasoning it presents than on seeing how that reasoning was produced.

Chatbots are already helping people bring claims

Vos also argues that courts must adapt to how people already seek legal help. He says people now bring legal problems to Gemini, ChatGPT or Claude, which can draft claim forms and particulars of claim—the written details setting out a case. He describes county court case starts as rising 10 percent annually and links that pressure to chatbot-assisted access.

How many of those AI-generated claims have a sound legal basis? Vos says the data is not available. He nevertheless rejects the assumption that a chatbot-written claim is automatically frivolous. Someone may misunderstand the precise nature of their problem, he argues, while still having a legitimate legal issue worth bringing to court.

For Vos, that is an access-to-justice argument, not merely an efficiency argument. He sees technology as a way to make a system historically available mainly to wealthy people reach more claimants. His optimism rests on the prospect of cheaper, faster resolution, while leaving the authority to permit AI judgments with the people affected and the wider public.