UK Ministry of Justice to trial AI for Crown Court transcription
TL;DR
- HM Courts & Tribunals Service will study whether its in-house Justice Transcribe AI tool can meet accuracy standards for Crown Court transcripts currently produced by external providers
- Transcripts can currently cost victims “hundreds or even thousands of pounds” — a barrier to court record access the study aims to reduce
- The work sits alongside the Victims and Courts Bill and a commitment to free sentencing-remark transcripts from spring 2027
The UK Ministry of Justice is launching a formal study into whether AI can handle Crown Court transcription at acceptable accuracy, led by HM Courts & Tribunals Service. The focus is the MoJ’s in-house tool, Justice Transcribe, which will be tested against the output of external commercial providers currently supplying transcripts — a service that for many victims is prohibitively expensive.
Why this use case matters
Courts and Legal Services Minister Sarah Sackman KC framed the rationale around victim access, noting that “victims show immense courage in coming to court” and that AI deployment could “boost transparency and access to justice.” The practical bar: transcripts that can run to four figures for victims seeking clarity on what happened in their own cases. Charlotte Schreurs, founder of the Open Justice For All campaign, welcomed the move, describing court transcripts as “imperative for victims in the healing process.”
The study fits into a wider modernisation push including the Victims and Courts Bill and a separate commitment that victims will have free access to transcripts of judges’ sentencing remarks from spring 2027. Justice Transcribe sits upstream of that commitment — if the AI tool meets accuracy standards, the economics of delivering free transcripts become dramatically more favourable.
The accuracy question is the whole question
Court transcription is a high-stakes domain for AI accuracy. Errors in a clinical note may need correcting; errors in a transcript of evidence can distort the legal record. The study’s central task is defining “acceptable” accuracy and establishing whether Justice Transcribe meets it under realistic conditions — multiple speakers, overlapping dialogue, accents, legal terminology, courtroom acoustics. Benchmark accuracy on clean audio is not the same as production accuracy on difficult trials.
This is also a procurement story. External transcript providers have an entrenched commercial position; in-sourcing transcription via AI changes the market materially. Expect push-back on methodology and validation standards from current suppliers.
Looking forward
The MoJ trial is a useful proxy for how UK government is approaching public-sector AI adoption: in-house tooling, defined use case, ministerial narrative built around a concrete user outcome (cheaper access for victims) rather than abstract productivity claims. If it delivers, Justice Transcribe becomes a template for other HMCTS workloads; if it fails on accuracy, expect a return to outsourced transcription but with pricing pressure reset. Either way, the study itself sends a clear signal that MoJ is willing to disrupt its own supply chain for access-to-justice gains.