What is a sovereign AI?
A sovereign AI is an AI system that you keep full control of: the data stays inside a perimeter you control (EU hosting), open source models run on your own infrastructure, and there is no dependency on a foreign vendor and no exposure to the US CLOUD Act.
How is this different from Mistral's Le Chat?
Le Chat by Mistral is a product that runs on Mistral's servers. We deploy and operate for you an infrastructure that you control, connected to your documents (RAG), with a daily wipe and integration with your SSO. You are not renting a service: you own your solution, and you can change model at any time.
Why not Albert, the French state's AI?
Albert (DINUM) is an excellent initiative, but it is aimed primarily at central government departments, and you need the in-house skills to operate it yourself. We provide a turnkey solution: deployed, maintained and supervised for you, with dedicated support, advanced document features (RAG, summaries, podcasts) and live references in local authorities.
Why not Microsoft Copilot or ChatGPT Enterprise?
These solutions send your data to servers subject to US law (CLOUD Act) and are billed per user, which becomes very expensive at scale. Our approach keeps the data in Europe, relies on reversible open source building blocks, and works at a fixed cost whatever the number of users.
Is it compliant with the GDPR and the EU AI Act?
Yes. The architecture is designed for compliance: data in the EU, structural minimisation (daily wipe), documentation provided for your DPO, pseudonymisation of personal data under review. We support you on the data protection impact assessment (DPIA) and on the transparency required by the EU AI Act. See also our AI & GDPR page.
How much does it cost?
The model is fixed cost: the infrastructure is pooled and billed for working hours only, so the bill does not grow with the number of users. Estimate your own case with our ROI simulator.
How long does it take to deploy?
Three weeks: one week of scoping and installation, then a two-week pilot with around twenty users, ending in a go/no-go milestone. You decide whether to roll out on the basis of measured results, with no blind commitment.
Where is the data hosted?
Exclusively in datacentres located in the European Union. User data (accounts, histories, knowledge bases) stays on a server you control; the AI processing servers only see the documents for the duration of the inference and are wiped daily.
Can we change language model?
Yes. Our stack is model agnostic: Mistral, Llama, Qwen or other open source LLMs, adapting to the GPU resources available. You are never locked into a single model. Details on our LLM models page.