AI providers management
How ATG selects AI providers based on your AI policy, runs its own European GPU infrastructure, and guarantees your data is never used for training.
Choose the right AI policy
The AI models used by ATG depend on your organization's AI policy setting.
- EU-only: only European providers are used
- Worldwide: EU providers plus selected non-EU providers
Learn more: AI Policy
Secure provider usage
All LLM interactions occur exclusively through APIs with contractual data protection guarantees:
- Your data is never used for model training — every provider contract includes strict non-training clauses
- No risk of data leakage via training or provider data collection
This ensures that your organization's knowledge remains confidential and is never exposed through AI model training or third-party data collection.
Provider scope
EU providers (data sovereignty guaranteed):
- Mistral - French AI leader
- Nebius - European AI infrastructure
- Black Forest Labs - German AI innovation
- Scaleway - French cloud & AI services
Worldwide providers
All EU providers plus:
- Cerebras
- Groq
- SambaNova
- Google Gemini
- OpenAI
- DeepInfra
Our own GPU infrastructure
On top of these providers, ATG also builds its own GPU infrastructure with Verda, a Finnish provider running data centers in Finland and Iceland.
- We deploy and operate some open-weights models ourselves, on GPUs dedicated to ATG
- We control the model version, its configuration, and its update cycle
- Inference stays within a European perimeter: Finland is in the European Union, Iceland is in the European Economic Area, and both are covered by the GDPR
These self-hosted models can be used under the EU-only policy, just like the European providers listed above. See Our approach to sovereign AI.