We asked eleven organisations who checks the numbers behind AI in law. Nobody has answered.

On 1 September this paper published a piece about conveyancing. It made a narrow claim and I was careful with it. We had looked for an independent measurement of what AI has done to the speed of legal work in Britain, and we had not found one. Three widely quoted numbers, three companies that sell the software.

The next day we wrote to all three, and to Sage, and to the Solicitors Regulation Authority, the Legal Services Board and the Insolvency Service. Over the following days we wrote to Uber, the GMB, the Information Commissioner's Office and HM Courts and Tribunals Service. Eleven letters. Every one to a published contact, every one from my own account, every one logged in our register before it went.

Seven of those deadlines have now passed. Nobody has replied.

I want to be fair about what that does and does not mean. The deadlines are ours. Seven to nine days is this paper's cadence, not a legal obligation, and nobody is required to meet it. Four Freedom of Information requests are inside their statutory windows and I am not counting them. Four more letters fall due today, and today is not over. Silence is not refusal, and an organisation that answers on Wednesday has done nothing wrong.

Something else happened in the same weeks.

On 5 August the SRA announced that applications had opened for the government's first Advisory AI Growth Lab. It is funded by government. It is designed to accelerate the development and deployment of AI products and services. Legal services was chosen as the first sector to get it, and applications close on 27 September. The SRA runs it alongside the Legal Services Board, the Council for Licensed Conveyancers and the Information Commissioner's Office.

Three of those four bodies are bodies we wrote to. None of them has replied.

So there is a government-funded programme to get AI products into legal services faster, run by the regulators, and those same regulators will not say whether anybody measures whether the products do what their suppliers say they do.

I am not going to inflate that into a scandal. It is a gap, and it is exactly the kind this paper exists to point at. You can be enthusiastic about the technology and still think somebody ought to be counting. Those two positions are not in tension. Accelerating something nobody is measuring is just going faster.

Ines has the machinery underneath it: how a figure gets made by a company that sells the thing, gets republished by the trade press, and turns up in your solicitor's marketing with nobody having checked it once. Elena has the four bodies, what each is for, and what we asked them.

One more thing, and it goes near the top rather than buried at the bottom. In that 1 September piece I described GenieAI's finding on entry-level hiring as a survey of its customers. It was not. GenieAI analysed job listings across Adzuna, Reed and LinkedIn, three of the largest boards in the country, and found paralegal vacancies down about 21 per cent over the year and more than 60 per cent since 2023. That is a stronger method than the one I named, and I named the weaker one while criticising the company. It is on the corrections page. It is the first thing I would want fixed if I were them.

The letters stand. We are still asking, and we will print whatever comes back.

— M.

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.