Security & governance

Advania ALT is designed for organisations that need to keep sensitive workloads on Swedish infrastructure while still using best-in-class models for less sensitive work.

Sensitivity tiers

Every request is classified by content sensitivity and routed to a model approved for that tier. See Models and sensitivity tiers.

Guardrails

Every request goes through the platform's guardrails before reaching a model. If a request cannot be routed to an approved model, the platform returns 403 with a decision id you can use to investigate. If the guardrail evaluation cannot complete safely, the platform returns 503 — it does not silently forward the request. See Guardrails contract.

Audit

The platform captures an audit log of significant platform events and a separate decision trail of guardrail routing decisions. Both are queryable via the console and the API. See Observability, cost & audit.

Regulatory posture

For questions about specific frameworks (EU AI Act, NIS2, GDPR, Schrems II, sector-specific requirements), contact your Advania Solutions Architect.