evidoria

← Back to browse

Good practice Imported

Chat Mistral — Luxembourg's Sovereign AI Platform for Civil Servants

Luxembourg · Luxembourg City · See the Luxembourg profile · See the Luxembourg City profile

Evidence: Descriptive / self-reported Top 51% 60/100 · Ask Evidence Copilot about this practice

In June 2025 Luxembourg's government signed a strategic partnership with Mistral AI to deploy 'Chat Mistral', a sovereign AI chatbot hosted in state data centres, giving the country's 36,000-strong civil service secure tools to search procedures and draft documents faster.

36,000
Civil servants with access to Chat Mistral
€10-12 million
Annual AI platform/infrastructure budget allocation (per year)
10 %
CTIE staff AI-training target (by end of 2025)
Chat Mistral — Luxembourg's Sovereign AI Platform for Civil Servants

Details

Maturity
Scaling
Promoter
Centre des Technologies de l'Information de l'État (CTIE), Government of Luxembourg, with Mistral AI
Period
2025–ongoing
Region (NUTS)
LU00
Keywords
public administration, internal operations, generative AI, data sovereignty

Context

On 17 June 2025, five members of Luxembourg's government -- including the Prime Minister and the ministers for the economy, defence, digitalisation and media/connectivity -- signed a multi-year strategic partnership with French AI company Mistral AI to build 'Chat Mistral', a large language model hosted entirely within Luxembourg's own government data centres.

Activities

The Centre des Technologies de l'Information de l'État (CTIE), which runs the programme, gives all 36,000 of the country's civil servants secure access to an AI assistant for searching procedural and HR documents and drafting routine text, has allocated roughly €10-12 million per year from its €175 million annual budget to AI platforms and infrastructure, and set a target of training 10% of CTIE staff on AI use by the end of 2025; all projects run through CTIE's 'Capital IT' methodology, which embeds security-by-design and the EU AI Act's risk classifications.

Results

As of the partnership's first months, no independent measurements of time saved, adoption rates or service quality were yet available.

Conclusions

The programme's case so far rests on Luxembourg's explicit data-sovereignty design -- all processing kept on state-controlled infrastructure -- and a concrete budget and workforce-training commitment, rather than on demonstrated productivity or service-quality results.

Implementation
Indicative cost
High (€500k–€5M) — CTIE has allocated roughly €10-12 million per year from its €175 million annual budget to AI platforms and infrastructure, a defined recurring cost line rather than a one-off grant.
Time to results
Medium (1–3 years) — Signed as a multi-year strategic partnership in June 2025, with companion tools (a Legilux legal chatbot, a citizen-facing guichet.lu integration) planned through end-2025 and beyond -- a phased, multi-year rollout.
Staffing & skills
Centre des Technologies de l'Information de l'État (CTIE), Mistral AI (technical partner)

Conditions for success

  • a formal multi-ministerial partnership (five ministers plus the Prime Minister) giving the programme political weight and a defined budget line
  • state-controlled data centres to keep all processing within Luxembourg's own infrastructure for data sovereignty
  • a defined methodology (CTIE's 'Capital IT') embedding security-by-design and EU AI Act risk classification from the outset

Common failure modes

  • no independent measurement of time saved, adoption or service quality exists yet, so the programme's value is currently asserted rather than demonstrated
  • the staff-training target covers only 10% of CTIE staff by end-2025, leaving most of the 36,000-strong civil service without confirmed training

Commonly funded by

National / regional programmes Digital Europe Programme

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful