Glossary

Air-gapped deployment

An air-gapped AI deployment is a system that runs with no network link to the outside world. Nothing goes in or out over a connection, so the software must do all of its work on site.

In a run

The agent gets its credential, the policy is checked and the record is written on the machine in your building, with no call to us or to any cloud. When the job ends, the record of every step is already on site.

Our word for it

Runs disconnected.

Which shape runs disconnected

Ares has three deployment shapes: cloud, hybrid and on-prem. Only the on-prem shape runs disconnected. It keeps working with no outside link, and the record stays in your building.

The check works the same way. Your administrators export the record. Your auditor runs the open verifier on a machine that is not ours, offline and with no account. If a record was changed, the check fails.

Before the site is disconnected, plan what each step needs. Every model, filter and policy the workflow uses has to be on the machine, because nothing can be fetched during a run. The record is written locally, and you export it for review in the way your program allows.

Where it stops

The cloud and hybrid shapes need a connection. A disconnected site also needs a local model for every step.

A program that must keep its data inside a boundary can run Ares this way. Read how Ares maps to public guidance, or compare the deployment shapes.

Related terms

  • On-prem AIOn-premises AI is AI that runs on hardware inside your own building instead of in a vendor’s cloud.
  • Local modelA local AI model is a model that runs on your own hardware rather than through a provider’s cloud service.

Tell us about one workflow.

We will say whether an agent should do it, and what the first step costs.

A senior member of the team replies within two business days.