Sovereign AI

Sovereign AI, European end to end

Opper on evroc runs the gateway on a Swedish cloud and routes to European-owned providers hosted in Europe, so your data stays with European companies from your app to the model.

See sovereign models

Your apps and agents

Your product

OpenAI SDK compatible

Your agents

Agent SDK

OpenAI-compatible API →

Opper on evroc

Stockholm, Sweden

Swedish company

Swedish cloud, evroc

DPA under Swedish law

ISO/IEC 27001:2022

European providers only

European providers

Sweden

4 providers

France

3 providers

United Kingdom

2 providers

Germany

1 provider

Ireland

1 provider

15 providers, 90 models →

How it works

European at every hop

The gateway, the cloud it runs on and the model provider are all European companies under European law.

Opper on evroc

The same Opper, running as its own instance at evroc.on.opper.ai on evroc's data centers in Sweden. Opper and evroc are both Swedish companies, so the gateway, its storage and your traces stay with European companies under European law.

  • Its own instance and account
  • OpenAI-compatible API
Deployment options 
import os
from openai import OpenAI
client = OpenAI(
    base_url="https://api.evroc.on.opper.ai/v3/compat",
    api_key=os.environ["OPPER_API_KEY"],
)
response = client.chat.completions.create(
    model="berget/zai-org/GLM-5.3-Flash",
    messages=[{"role": "user", "content": "Summarise this contract"}],
)

European providers, set once

Call any route on the sovereign models list by its model id. On Control Plane, set one Model access rule with provider country and inference location in Europe and it becomes the default for your whole organization. Each route shows its retention and training posture, and many European routes carry zero data retention.

  • Provider country and inference location in one rule
  • No silent switch to another provider
How Model access rules work
Model accessDocs ↗
provider countrySE FR GB DE IE IT LU NL ES
inferenceEEA · the EU, Norway, Iceland, Liechtenstein
outside the rulethe call fails · no silent switch

Rules for the whole organization

On Control Plane you set five rules once for the whole organization: a monthly spend limit, how long traces are kept, which models and providers are allowed, checks that redact or block sensitive content before a model sees it, and the order providers are tried in.

  • Spend limits per organization, project, user or key
  • Traces kept 1 to 30 days, or not at all
See the AI control plane
RulesDocs ↗
Spend limits$2,000 / month · caps per project
Data retention30 days · or nothing stored
Model accessEU · zero data retention
ChecksPII · redact · secrets · block
Routingcheapest first · failover in the pool

Live data

European-owned inference providers

Every provider headquartered in Europe with models hosted in Europe, straight from the model directory.

ProviderHeadquartersZDRModels in Europe
MistralFranceYes24
GreenPTNetherlandsYes18
BergetSwedenYes14
evrocSwedenYes14
TensorXIrelandYes14
GreenferenceFranceYes12
RegoloItalyYes12
ScalewayFranceAbuse monitoring11
InceptronSwedenYes7
SferenceUnited KingdomYes7
MeliousGermanyYes6
NextbitSpainYes5
OpperSwedenYes5
InfercomLuxembourgYes3
SLNGUnited KingdomYes2

15 providers in 9 countries serve 90 models from Europe today, on 154 routes. Headquarters is the provider company's home country; UK and Swiss companies are included, so set provider country in a Model access rule if your policy draws the line at the EU or the EEA. Route-level retention and training posture is on each provider page, and the sub-processor register is on the trust center.

FAQ

Sovereign AI FAQ

What is sovereign AI?

+
AI where the companies that handle your data are European and answer to European law. For an AI gateway that means two hops: the gateway that receives your prompt, and the model provider that runs it. Opper on evroc keeps both with European companies when you call European-owned routes.

How is Opper on evroc different from opper.ai?

+
It is the same Opper, running as its own instance on evroc, the Swedish cloud, at evroc.on.opper.ai, with its own account, API keys and API address. The main Opper platform runs in AWS Stockholm. Projects build against the same API on both, so you pick the instance that matches your policy.

How do I get access to Opper on evroc?

+
Opper on evroc is set up with your team rather than by self-service sign-up. Get in touch and we onboard your organization, with its own account and API keys on the evroc instance.

Which models can I use on the sovereign path?

+
Every model from a provider headquartered in Europe and hosted in Europe: Mistral's own models, and open-weight models on European hosts such as Berget, evroc, GreenPT, Regolo and Scaleway. The provider table on this page lists every European-owned provider with its live model count.

Can I use Claude, GPT or Gemini?

+
Yes, on the main Opper gateway, where they run from EU data centers on AWS, Azure and Google Cloud. That covers EU residency and the GDPR, which is what most teams need. Those clouds are US-owned, so for a policy that requires European ownership end to end, use the European-owned routes on Opper on evroc.

How do I keep every call sovereign?

+
Call any route on the sovereign models list by its model id. On Control Plane, set one Model access rule with provider country and inference location in Europe and it becomes the default for your whole organization: a call to anything else fails with a clear error, and Opper never switches to another provider silently.

Is Opper on evroc GDPR compliant?

+
Yes. Opper Technology AB, a Swedish company, is your processor under Opper's published Data Processing Agreement, and evroc is listed on the Opper trust center as one of Opper's two production environments. On the sovereign path, the gateway and the model provider are both European companies.

Residency

Claude, GPT and Gemini hosted in the EU

Compliance

The GDPR-compliant AI gateway

Guide

How to use LLMs under the GDPR

Run sovereign AI on Opper

Opper on evroc is set up with your team. Get in touch and we onboard your organization.

See sovereign models