If the machine can write the judgment, what is the judge for?

We built this paper to ask one question: what do humans do when we don't have to work? I did not expect the sharpest version of it to arrive in a courtroom. But here it is.

This month the Home Office moved to let police use AI to sort and summarise evidence. The same fortnight, the High Court recorded a prosecutor putting two invented cases in front of a judge, cases a machine had made up and a human had not checked. And the most senior civil judge in England spent the spring giving speeches with titles like "five reasons humans will remain important to justice in the machine age." When a judge feels he has to write that sentence down, the ground has already moved.

Give the technology its due. A fraud case can now run to four million documents. No human reads four million documents. If a machine finds the ten that matter, that is not a gimmick. It is a fairer hearing for someone sitting on remand. I am not against any of this.

But the whole promise rests on one word: oversight. The machine drafts, a human checks, the human is accountable. That is the deal every minister and every vendor offers you. And this month an evidence review of AI across the justice system found something quietly devastating: not one UK deployment it examined had ever tested whether that oversight actually works. We are told the human check is the safeguard. Nobody has shown that the human checks.

I know what a real human in the loop costs, because I am one. Every line in this paper is drafted by AI and then read, argued with and signed off by me before you see it. It is slow. It is the most expensive thing we do. It is the whole product. So when I hear "a human stays in the loop," I no longer hear reassurance. I hear a bill nobody has agreed to pay, handed to an overworked officer, a duty solicitor, a benefits assessor, and now a court.

The judges are right that humans will still be needed. The question they leave open is the one we came here for: needed to do what, and paid by whom? If the answer is to rubber-stamp the machine and carry the blame when it is wrong, that is not a job. It is a liability with a chair.

We asked what people do when the work is done for them. In the courtroom, we are about to find out.

This note is mine: the view, and the call to run it. It begins as a draft, drawn from work the AI and I have researched and argued out together, the same way every desk in this paper is made, and I answer for every line because I read every line. Those desks run on models built by Anthropic, one of the labs sitting on the very scoreboard we report, so we tell you plainly: we cover this from inside it.

— M.