---
title: "UK justice system publishes the number of its AI summaries, not their error rate — Marsanne Studio AI watch"
description: "The UK Ministry of Justice publishes usage volumes for its AI tool Justice Transcribe, with no data on its reliability."
url: "https://marsanne.studio/en/ai-watch/uk-justice-system-publishes-the-number-of-its-ai-summaries-not-their-d6d02a/"
lang: "en"
---

[Marsanne Studio](https://marsanne.studio/en/)

[← Back to home](https://marsanne.studio/en/ai-watch)

[← The AI watch](https://marsanne.studio/en/ai-watch)

Analysis · 2 October 2026

# UK justice system publishes the number of its AI summaries, not their error rate

Original title, in French: “La justice britannique publie le nombre de ses résumés d'IA, pas leur taux d'erreur”

**Publication**

ActuIA (France)

**Author**

Stephane Nachez

**Published on**

2 October 2026 · read in the page structured data · updated on 2 October 2026

## Summary

On 24 September 2026, the UK Ministry of Justice published the review of the first year of its AI plan, adopted in July 2025. The Justice Transcribe tool, which summarises probation meetings, has processed more than 1.6 million meetings, up from 150,000 in February, according to the four published statements. An internal department estimates an illustrative saving of 266,667 hours, based on an assumption of ten minutes per meeting. No public document gives a transcription error rate, despite a 50% reduction in administrative time cited by ministers before Parliament without a documented method. In France, the Ministry of Justice rolled out "Mon Assistant Justice" in May 2026, without publishing a comparable usage volume.

Summary based on the full text of the article.

[Read the article on ActuIA →](https://www.actuia.com/actualite/la-justice-britannique-publie-le-nombre-de-ses-resumes-dia-pas-leur-taux-derreur/)

## Figures cited

-   1,600,000 meetings
    
    Number of probation meetings summarised by Justice Transcribe as of 14 September 2026
    
    United Kingdom, 7 October 2025 to 14 September 2026
    
    > « Plus de 1 600 000 au 14 septembre. »
    
    Source: Ministère britannique de la Justice
    
-   1,500,000 meetings
    
    Probation meetings summarised according to the ministry's review
    
    United Kingdom, 7 October 2025 to 14 September 2026
    
    > « « plus de 1,5 million » d'entretiens résumés entre le 7 octobre 2025 et le 14 septembre 2026 »
    
    Source: Ministère britannique de la Justice
    
-   266,667 hours
    
    Estimated hours saved thanks to Justice Transcribe
    
    United Kingdom, illustrative estimate
    
    > « ce qui donne « environ 266 667 heures » »
    
    Source: Probation Workforce Transformation, HMPPS
    
-   50 %
    
    Reduction in administrative note-taking time according to a minister
    
    United Kingdom, parliamentary statement of 3 December 2025
    
    > « « cet outil d'IA de pointe a réduit d'environ 50 % le temps administratif de prise de notes, avec des scores de satisfaction remarquables » »
    
    Source: Lord Timpson, ministre d'État à la Justice
    
-   4.7 out of 5
    
    Average user satisfaction rating for the tool
    
    United Kingdom, statement of 2 June 2026, more than 63,000 reviews
    
    > « plus de 63 000 avis déposés dans l'application, note moyenne de 4,7 sur 5 »
    
    Source: Ministère britannique de la Justice
    
-   2,500 staff
    
    Staff who took part in the Mon Assistant Justice pilot
    
    France, before the rollout of 12 May 2026
    
    > « après une expérimentation menée avec 2 500 agents »
    
    Source: Ministère français de la Justice
    

## Classification

**Country**

United Kingdom, France

**Company size**

Large companies

**Industries**

Government and local authorities

**Companies**

DINUM (supplier), Ministère britannique de la Justice (deploys AI), Ministère français de la Justice (deploys AI), HMPPS (deploys AI)

**Keywords**

generative artificial intelligence, justice, transcription, algorithmic transparency, probation, AI regulation

Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to ActuIA and its author. Entry translated from French by Claude Sonnet 5.5.
