Loading the catalogue…
Loading the catalogue…
No “trust us.” Here’s exactly how your data is handled, where it runs, who can touch it, and which frameworks we’re built to — the answers your security and compliance teams ask for first.
Prompts and completions are never used to train models — ours or a provider’s. Inputs are processed and discarded per your retention setting.
TLS in transit, encryption at rest. Provider keys are stored write-only with envelope encryption — never logged, never returned.
Request and routing logs follow a tier you choose — 7 days to 7 years. Prompt content can be excluded from logs entirely.
Each request is logged with model, provider, country and jurisdiction — the evidence trail, not just the metrics.
We route only to providers we’ve verified run under EU law. Commercial frontier models are only ever called when your sovereignty policy permits, and every such call is flagged in the audit log.
| Role | Location | Jurisdiction | |
|---|---|---|---|
| Scaleway | Sovereign model hosting | Paris · Amsterdam | EU (FR/NL) |
| OVHcloud | Sovereign model hosting | Gravelines · Frankfurt | EU (FR/DE) |
| STACKIT | Sovereign model hosting |
Aligned today, with formal certifications on the roadmap as we scale. Provider-level attestations (e.g. SecNumCloud, BSI C5) are inherited from our vetted sub-processors. Stav provides compliance features and evidence — your DPO makes the determination.
Aligned
Ready
Exempt by design
Aware
Aware
In progress
Roadmap
Mapped
Sub-processor list, data-flow diagrams, DPA and the audit export — everything a review needs, in one place.
| EU (DE) |
| Mistral | First-party model hosting | Paris | EU (FR) |
| IONOS | Sovereign model hosting | Germany | EU (DE) |
| Commercial model APIs | Routed only when you allow | EU endpoints / BYOK | Policy-gated |