Skip to content
Marsanne Studio
← The AI watch

Analysis ·

Mistral AI in government ministries: what the State is buying for €6m

Original title, in French: “Mistral AI dans les ministères : ce que l'État achète pour 6 M€”

Publication
ActuIA (France)
Author
Stephane Nachez
Published on
7 September 2026 · read in the page structured data · updated on 7 September 2026

Summary

Les Échos revealed on 2 September a new €6m contract for 2026-2027 between the French State and Mistral AI, confirmed by the Ministry of Public Action and Accounts. The non-exclusive agreement provides for Mistral engineers to be seconded to ministries to develop AI use cases in cybersecurity, fraud prevention and the handling of mass litigation in the Justice ministry. It extends the rollout of L'Assistant, DINUM's conversational tool, made generally available in June after tests with 10,000 civil servants (75% find it useful, with time savings of up to 16%). The contract forms part of a €200m cyber plan announced in April, following a data leak at the DGFiP affecting 678,000 people. Minister Roland Lescure nevertheless warned against exclusive dependence on Mistral for European AI.

Summary based on the full text of the article.

Read the article on ActuIA →

Figures cited

  • 6 M€

    Value of the contract between the State and Mistral AI for 2026-2027

    French ministries, 2026-2027 period

    « six millions d'euros sur la période 2026-2027, un contrat sans caractère exclusif »

    Source: Les Échos / ministère de l'Action et des Comptes publics

  • 200 M€

    Budget of the State cyber plan announced in April

    French State, funded from France 2030

    « Sébastien Lecornu annonçait 200 millions d'euros prélevés sur France 2030 »

    Source: Matignon

  • 5 %

    Share of its digital budget that each ministry will have to devote to cyber from 2027

    French ministries, from 2027

    « Chaque ministère devra consacrer 5 % de son budget numérique au cyber dès 2027. »

    Source: Matignon

  • 10,000 civil servants

    Number of civil servants who tested L'Assistant

    Six pilot ministries, 10 months of trials before June 2026

    « L'outil sortait de dix mois d'expérimentation auprès de 10 000 agents dans six ministères pilotes. »

    Source: DITP

  • 75 %

    Share of L'Assistant testers who find it useful for their work

    L'Assistant test users, pilot ministries

    « Le bilan de la DITP retient que 75 % des testeurs le jugent utile à leur métier, que 65 % le recommanderaient à un collègue, et un gain de temps mesuré jusqu'à 16 % sur la synthèse documentaire. »

    Source: DITP

  • 65 %

    Share of testers who would recommend L'Assistant to a colleague

    L'Assistant test users, pilot ministries

    « Le bilan de la DITP retient que 75 % des testeurs le jugent utile à leur métier, que 65 % le recommanderaient à un collègue, et un gain de temps mesuré jusqu'à 16 % sur la synthèse documentaire. »

    Source: DITP

  • 16 %

    Measured time savings on document summarisation thanks to L'Assistant

    L'Assistant test users, pilot ministries

    « Le bilan de la DITP retient que 75 % des testeurs le jugent utile à leur métier, que 65 % le recommanderaient à un collègue, et un gain de temps mesuré jusqu'à 16 % sur la synthèse documentaire. »

    Source: DITP

  • 2,500 civil servants

    Number of Mon Assistant Justice testers before wider rollout

    Ministry of Justice, before 12 May 2026

    « après un test auprès de 2 500 d'entre eux »
  • 678,000 people/entities

    Number of individuals and professionals affected by the DGFiP data leak

    DGFiP, unauthorised access in June-July 2026

    « avaient servi à consulter et extraire des données fiscales et cadastrales concernant 678 000 particuliers et professionnels »

    Source: DGFiP

Classification

Country
France
Company size
All sizes
Industries
Government and local authorities
Companies
DINUM (deploys AI), DGFiP (cited), Outscale (supplier), ministère des Armées (deploys AI), Assurance Maladie (deploys AI), Mistral AI (supplier), France Travail (deploys AI)
Keywords
public sector, cybersecurity, fraud prevention, mass litigation, conversational assistant, public procurement

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.