← Back to the engine

Defence · later

AI in defence without classified information leaving the network.

Detector not built yet

In defence the question is not only who, but which classification. The same engine can get a detector that knows the markings, national and NATO's, and enforces that nothing above a given level reaches a model that must not see it. The detector is not built, and we work with public material only.

What Suvard would protect here

A classification-aware detector: the markings as the ontology.

National classification markings
FOR OFFICIAL USE, CONFIDENTIAL, SECRET
Not built
NATO markings
NATO RESTRICTED and above
Not built
Units, positions and equipment IDs
names and identifiers
Not built
Personal data about staff
GDPR
Not built

The defence detector is not built, and we work with public material only. The idea is the same engine with a new detector: the marking decides which model and which network a request may reach, and the seal proves afterwards that the rule was followed.

The frameworks it would be built for

Supervision: The Danish Ministry of Defence and the Centre for Cyber Security

Classification rules

national and NATO's: the marking decides the handling

EU AI Act

exempts purely military purposes (art. 2); civilian administrative uses are covered

GDPR

staff personal data

Suvard provides the technical controls and the evidence; the legal assessment is yours. Not legal advice.

Where it would be used

  • Lessons learned and report work on public material
  • Administrative drafts without personal data in the request
  • Case handling where the marking decides which model may answer

Other sectors

Next step

A 30-minute check on where AI is already running in your organization.

We start with a mapping, not a sale. You get an honest picture of what passes through your network, and what it would take to be able to prove it afterwards.