Mistral and sovereign AI: when to favour it for your clients
Announced EU hosting, GDPR, buyer requirements: when to offer a client the Mistral Small option, and what you should be careful not to promise them about it.
Security and data5 min read
“Where does our data go?” It is one of the first questions a client asks when you propose an AI assistant. For some it is mere curiosity. For others, such as a local authority, a medical practice or a law firm, it is a condition for signing.
blabla.bot offers two model families: GPT-6 Luna from OpenAI and Mistral Small from Mistral AI. We present Mistral Small as the sovereign option. This article explains exactly what that means, when to offer it, and what not to promise your client.
What we mean by “sovereign”
The word is used in many different ways. In the context of blabla.bot it means one precise thing: the model provider states that it processes data in the European Union.
Mistral AI says in its help centre that data is hosted in the European Union by default. It also mentions two limits:
- a customer can explicitly choose a US endpoint, in which case their data is hosted in the United States;
- depending on the feature, data may be temporarily transferred outside the European Union, to subprocessors listed in its Trust Center.
blabla.bot has not verified this hosting itself. That is why the console says “announced EU hosting”, and it is the wording we recommend you use with your clients.
What the GDPR says, and what it does not
A common belief is that the GDPR forbids sending personal data outside the European Union. That is not accurate. The regulation governs such transfers: they are allowed to a country the European Commission recognises as offering adequate protection, or with appropriate safeguards such as standard contractual clauses.
The French data protection authority (CNIL) notes, for example, that the European Commission adopted an adequacy decision for the United States in July 2023, which allows certain transfers to US organisations.
In other words, choosing Mistral Small is not a legal requirement for a website assistant. It is a cautious choice or a client requirement. Whether an assistant complies with the GDPR depends on many other points: informing visitors, retention periods, the data collected, the processing agreement. A separate article covers that topic.
When to favour the Mistral Small option
These are the situations where the sovereign option is most often justified.
| Situation | Why |
|---|---|
| The client requires it in a specification or tender | Some public buyers and large accounts set conditions on where processing takes place. |
| The sector handles sensitive information | Health, legal, social services: even if the assistant should not receive sensitive data, visitors sometimes type it anyway. |
| The client’s internal policy favours European providers | A legitimate preference, and it eases the conversation with their data protection officer. |
| The client asks and the price difference suits them | An explicit choice is better than a doubt that will come back later. |
Otherwise, the default recommended option, GPT-6 Luna · Simple, remains a good starting point.
The cost difference
The sovereign option costs a little more. With the catalogue prices checked on 13 November 2026 and the console’s conversation assumptions, a conversation is estimated at:
- €0.0042 with GPT-6 Luna · Simple;
- €0.0060 with Mistral Small · Simple.
That is about 40% more, or less than two euros per 1,000 conversations. For most sites you can absorb that into your price without difficulty. Run the numbers at your volume before raising it with the client.
What the sovereign option does not cover
Choosing Mistral Small concerns the model that writes the answers. To be straight with your client, also make the following points clear.
Knowledge search. To find the right passages, blabla.bot turns the client’s content into vectors (embeddings). That service is provided by Mistral AI for every agent, whatever answer option is chosen. So a client who picks Mistral for answers does not see their indexed content go to another provider for search.
The rest of the infrastructure. A conversation also goes through blabla.bot’s servers and database. The region policy for the platform as a whole is still being defined: we do not promise today that every processing step stays in the European Union.
Compliance. No model choice makes an assistant “GDPR compliant” on its own.
How to discuss it with your client
A simple, accurate statement beats a fragile sales pitch. For example:
“We can run your assistant on Mistral Small, a model whose publisher states that it hosts data in the European Union by default. It costs a little more. On its own, this choice does not guarantee that the assistant complies with the GDPR: we will go through the other points together.”
Then, in the console, select the Mistral Small option on the agent’s “AI model” screen. The change applies from the next answer. Test the agent in the simulator to check that the answer quality suits you.
In short
The sovereign option meets a real requirement for some clients. Offer it when the client asks, when their sector calls for it, or when the price difference is not an issue. And present it for what it is: announced European hosting for the answer model, not a compliance guarantee.
You will find more answers about data and security in the FAQ.
Sources
Tags
- sovereignty
- Mistral
- GDPR
- data
On the site
Read next
Costs and AI models
Choosing your AI model: quick answers or in-depth answers
Simple or In-depth, GPT-6 Luna or Mistral Small: what really changes between the blabla.bot AI options, and how to choose the right one for each client site.
· 5 min read
Security and data
AI chatbots and the GDPR: what to check before going live
Informing visitors, sensitive data, retention, trackers, processors: the GDPR and AI Act points to check before you publish an AI chatbot on a client website.
· 6 min read
Costs and AI models
How much does a conversation with an AI assistant cost? Tokens explained
Input tokens, output tokens, context and history: what really makes up the cost of a conversation with an AI assistant, and how to bring that cost down.
· 5 min read