THE MEDIA FOR AI THAT ACTUALLY HELPS.English edition · Understand & take action
Spotted on the web

Transmit judicial hearings twice as fast

Use described publicly by IA dans l’État — DINUM (accessed in November 2026 (undated page)), summarised by the desk. Not tested by iadujour.

The result to aim for

The Digital Agency of the Internal Security Forces has made available to investigators PAROLE, a platform for automatic transcription of court hearings and recordings used by 6,500 people.

What was done

The fact sheet on PAROLE recalls that 85,000 hours of court hearings are conducted each year and that one hour of hearing is traditionally six to seven hours of transcript. Designed by NATO as a common solution for the internal security forces, the platform primarily serves the filmed hearings of minors victims and judicial listening. The investigator files the audio or video file; The tool transcribes it in a few minutes, distinguishes the speakers, then an integrated editor allows to correct, to reallocate the words and to produce a workable text in the procedure, synchronized with the video. According to the page, one hour of recording is processed in six to seven minutes, 4,000 hours of audio spend each month on the platform, and the transcript work is divided by two (three to four hours per hearing hour). The architecture is fully internal, with unique traceability and authentication, and has ended unframed local initiatives. The fact sheet states that the pilot project continues on a voluntary limited perimeter, with a gradual spread.

Before you start

Tool cited by the source: PAROLE (ANFSI). Fill in the bracketed items: [coller]. Strip confidential data that is not needed and check your tool’s access terms. Features and plans can change.

The method, in three steps.

  1. Gather the context

    Spot the bracketed items in the prompt ([coller]) and prepare what they ask for.

  2. Adapt the prompt

    Copy the prompt below, replace the bracketed items, then paste it into the AI assistant you use. The cited tool (PAROLE (ANFSI)) is not necessarily available to the public..

  3. Reread before using

    Reread the result, correct it if needed and keep the final decision. Caveats are detailed below.

Your starting point.

Replace the bracketed items, then adapt the proposal to your situation. This prompt is written by the iadujour desk It is not a quotation from the source.

The prompt to customise
Here is an automatic interview transcript, with its speakers: [coller].

Help me read it again without rewriting it:
1. Report passages where the speaker's attribution seems doubtful.
2. Finds the probably poorly recognized words (own names, places, numbers, dates) and proposes a hypothesis in square brackets.
3. Indicates inaudible or cut passages.
Never change the meaning or wording of the words.

What deserves your attention

Human correction remains a mandatory step: an error of transcription in a judicial procedure may change the meaning of a declaration.
Origin of the case. Use described publicly by IA dans l’État — DINUM (accessed in November 2026 (undated page)).
Proof level: Institutional source. Source checked on 10 November 2026. Use described by the source, not tested by the desk. Original summary written by iadujour.fr: figures and claims are the source’s; the prompt is a desk proposal, not a quotation.
TO GO FURTHER

Other ideas to put to work.

The whole Legal theme
Legal

Build yourself the small tools of a legal team

Without a developer, the legal team of Anthropic has created a "standard" with Claude Code that directs to the right lawyer and tools for monitoring legal journals.

Tool: Claude CodeType: Spotted
DISCUSSION

Comments.

A field note, a question, a precision on this use case? Comments are reviewed before publication.

Published comments

Loading comments…

    Leave a comment

    Moderation charter. Every comment is reviewed before publication. We refuse hate speech, insulting or discriminatory language, advertising, spam and promotional links, off-topic posts and other people’s personal data.