aigw
Sovereignty

An AI gateway built for sovereignty

Wherever you operate, the questions about AI are the same. Where does the prompt go, who holds the keys, and can you prove it? aigw is built so the answers are yours to give.

The problem

Most AI traffic goes wherever the provider’s default endpoint is, and the data in a prompt is often the most sensitive thing an organization has. Data protection and residency rules in the EU, Canada, Australia, India, Japan and elsewhere keep asking the same three things: where the data is processed, who can reach it, and what record shows it. A gateway run by someone else, in someone else’s jurisdiction, answers none of them for you.

How aigw keeps it in your hands

  • Run the gateway in your own region. The aigw gateway runs in your own cloud account, data center, or Kubernetes cluster, in the country you choose. Prompts and responses pass from your applications through your gateway to the model provider, and never through us. The gateway needs one outbound HTTPS connection to the control plane and nothing inbound.
  • Choose where each model runs. Route every model to a provider and region you pick, including providers you host yourself, so traffic stays where your rules require.
  • Hold your own keys. aigw runs on your own provider accounts. Keys are sealed at rest; the gateway can use them, and the applications and agents calling through it never see them.
  • Show the evidence. Every call gets a governance record of who called what, which policy applied, and what was blocked. Every configuration change lands in a tamper-evident audit log. Both can go to your own observability stack over OpenTelemetry.

Where the control plane runs

The managed control plane is hosted in the EU. It holds configuration, policy, and the metadata of each call (model, token counts, cost, latency, status, and governance flags). With a self-hosted gateway it never receives prompt or response content. If your rules require the control plane in your own country as well, talk to us.

Who it is for

  • Public sector and regulated industries that must keep AI processing in a given jurisdiction.
  • Companies in markets where the large providers’ default endpoints are not an option, and that need to route to the providers available locally.
  • Platform teams that want one governed endpoint for every model, tool and agent, without handing their keys or their traffic to a third party.

See how it works for the EU in Keep AI in the EU, and how to run your own gateway in self-hosted gateways.