Private AI: your data, in Europe
Your agents and their files live on servers in the European Union, each one in its own isolated space. On the Professional plan we sign the GDPR data processing agreement and you get European AI models that don't store your data or use it for training.
Where it is, who sees it and how long it is kept
Where it lives
All the infrastructure is in the EU: Helsinki (Finland), Nuremberg and Falkenstein (Germany). The provider is Hetzner.
Who sees it
Nobody accesses your content except to provide the service or to resolve an issue you ask us about.
What we keep
For billing we keep metadata — model, tokens, cost — never the text of your conversations.
For how long
Your data, for as long as your agent exists. Daily backups expire in 56 days at most, and the immutable backup in 7.
Who is responsible for what, depending on your plan
For your own personal use
- We are the data controller only for your account data
- What you process with your agent is your responsibility
- No data processing agreement
- Standard models
For customer data
- We are your data processor (GDPR art. 28)
- Processing agreement and list of sub-processors included
- Full catalogue, including the European (EU) models
- We tell you before we change sub-processor
If you process data about customers, employees or suppliers: Professional.
Which model to use, depending on the data you give it
The European models carry the (EU) tag in their name. You choose which models each agent uses from your account area.
| Models | Where they process | What they guarantee | Use them for |
|---|---|---|---|
| European (EU) | |||
| Mistral | EU and EEA | Processing agreement, zero retention and training switched off | Everything, including personal data |
| Scaleway | Paris (France) | Doesn't keep requests or responses and doesn't train on them | Everything, including personal data |
| Non-European | |||
| OpenAI, Anthropic, Google, xAI and others | United States and other countries | The most capable models on the market | Coding, analysis, research or content, with no personal data |
| Experimental | |||
| Qwen free | Germany and Finland | No processing agreement | Testing with no personal data |
Professional plan rule: when a task involves personal data of people living in the EU, the models tagged (EU) are used. Details in the data protection information.
How we protect it
Every agent in its own space
Each agent runs in its own container, with separate storage.
Encrypted communications
Connections to the model providers are encrypted with TLS.
Backups against ransomware
Verified daily backups and a separate immutable backup.
Your permissions, your rules
Every connection — email, Drive, Notion — with read or write access, and you remove it whenever you like.
Change log
Configuration changes leave an audit trail.
Privacy isn't about nice phrases: it's about judgement
Connecting everything at once, giving write access to everything and pasting passwords into the chat.
Minimum permissions, start with read-only, remove what you don't use and use the (EU) models when personal data is involved.
Privacy isn't doing nothing. It's connecting what you need, with just the right permissions.
What people ask about privacy and the GDPR
What is a private AI?
Where is my data?
Does it comply with the GDPR?
Do you use my conversations to train models?
Can I ask you to delete my data?
Can I use free ChatGPT with customer data?
Private AI for your business, from €19.99 a month
Professional plan · with a data processing agreement
- Servers in the European Union
- European (EU) models with zero retention
- Every agent, isolated
- No lock-in