Skip to main content
Fit to CareCare into Action

The Record

Articles and public records from Fit to Care, written from checked evidence. Every piece names its sources. Where a publication ran a piece first, we say so and link to it. Nothing here was paid for by any party, person or client.

No record is perfect. If you find an error, write to anthony.lawton@ffmi.co.uk with the subject "The Record correction" and it is corrected within 48 hours, in public, on the piece itself.

Article

Did the Prime Minister answer? On the same 27 answers, two AI readers agreed 93% of the time; pairs of people agreed 30% to 63%

Can AI help people agree on whether a question at Prime Minister's Questions was answered? Two AI readers agreed with each other on 82% of 9,480 answers. Pairs of human markers, checking a sample, agreed on between 30% and 63%. When one marker read the machine's reasons, he changed his mind on 7 of the 11 differences he looked at again.

Parliament, Prime Minister's Questions

Article

Whoever governs, three in four opposition questions go without a full answer: ten years of PMQs

Ten years of Prime Minister's Questions: only about one question in three got a full answer. When their party governed, Conservative and Labour MPs went without a full answer about four times in ten; in opposition, about three times in four. Personal remarks made little clear difference to whether a question was answered, but went with replies that attacked the questioner.

Parliament, Prime Minister's Questions

Working paper

The Failure Modes Public Bodies Are Being Instructed to Build

What a public body still cannot catch when it follows UK government AI guidance exactly as written. Three recurring patterns, drawn from the guidance itself and the audit record of the National Audit Office and the Public Accounts Committee. Working paper v1.2, August 2026, revised 8 September 2026.

AI and public spending, NHS, Schools, Councils