Private EU-hosted AI
Secure AI for business, with a private model that is yours
Secure AI for business means you decide where your data goes. In Managed Cowork you work with a private model, DeepSeek or Mistral, in a shielded environment that is truly yours, within the EU. For general work you add an online model alongside it.
- A private DeepSeek or Mistral model
- Sensitive data stays within the EU
- Every question and answer is logged
Where your data goes
Private AI: your data does not go to another party
Secure AI in your business starts with knowing what happens to your data. With the private model in Cowork the answer is simple: the language model runs in its own shielded environment within the EU, and what you put in does not go to another party. Your data is not used to train models either.
Your files and conversations are stored encrypted in the Infodatek Group data centre in the European Union, on our own hardware. Every conversation gets its own closed workspace, and the keys of AI providers stay on our server, never in your browser.
Digital Care runs at Infodatek Group, certified to ISO 27001 and NEN 7510. We set up the environment, arrange the permissions and keep everything up to date.
The right model for the task
AI within the EU for sensitive work, online models for the rest
Not every question is equally sensitive. That is why Cowork switches between language models depending on the data you put in. Working on an HR file or a client contract? That stays with the private model within the EU. Rewriting a text or looking something up is done with the online model that suits it best: ChatGPT, Claude, Copilot, Gemini or Grok.
We arrange the licences with those providers. All usage comes from one AI budget for the whole organisation, on one invoice.
- Set per user which models are available
- Combine the private model with online models
- New models without a new contract
Ground rules
Fixed ground rules for the AI, whatever the model
A secure model is one thing, clear agreements are another. In Cowork the same rules apply to the AI, whichever model you choose. A person stays in control, and you can always trace where an answer came from.
- The AI only draws on sources the user may access
- Every answer shows its source, so you can check it
- Your data trains no model at all
- A suggestion only goes ahead once a person approves
- Questions and answers are kept in the log
Permissions per role
What you get back depends on who you are
There is an extra check between your systems and the AI. It applies the permissions of the person asking before any data goes back. Per integration and per role you decide what is allowed: nothing, read, write or manage. Move to another model later, and your permissions do not change.
Signing in is done with your own Microsoft 365 account. Your organisation's security, such as two-step sign-in, applies here too.
- Integrations are only on when you switch them on
- Writing to your systems is off by default
- Every change in permissions goes into the audit log
Approach
How to start working securely with AI
- 1
Get acquainted
In a live demo we show how the private model and the online models work together, based on your situation.
- 2
Choose models and roles
Together we decide which models are available and which role may use which sources.
- 3
Set up
We set up the shielded environment and connect the systems you want to use.
- 4
Get going
Everyone signs in with their own work account. The log shows what was asked and answered.
Frequently asked questions
Questions about secure AI for business
Where does my data stay when I use AI?
Your files and conversations are stored encrypted in the Infodatek Group data centre in the European Union. When you work with the private model, the AI stays within the EU as well, in a shielded environment that is yours. If you choose an online model for general work, your question goes to that provider.
Is my business data used to train AI?
No. Not using your data to train models is a fixed agreement in Cowork, for every model. On top of that, every question and answer is logged, so you can always see what was done with the AI.
Which private AI model can I choose?
You choose between the DeepSeek and the Mistral language model. They run in a shielded environment within the EU, and you can combine them with online models. Who may use which model is set with the permissions per role.
Can I use online AI models alongside the private model?
Yes, in the same workplace. We arrange the licence per user with the provider, and permissions per role and the log apply there just the same. Keep sensitive data with the private model, and use the online model that fits for general work.
How do I show that we use AI responsibly?
Every question and answer is in the log, with source, user and time, and changes to permissions go into the audit log. Digital Care also runs at Infodatek Group, certified to ISO 27001 and NEN 7510. You can point your own clients to that as well.
See Digital Care in your own environment
A half-hour conversation with one of our specialists. Your own systems, your own questions. No sales pitch.