---
title: Air-Gapped Computing
slug: air-gapped-computing
qid: local:air-gapped-computing
snapshot: air-gapped-computing-000012-dd8534300ecc645e
content_hash: dd8534300ecc645eb32e0874d4e4d4e9a94500a91ca342286978b4d0d53010d1
license: Anticloud Synthetic Licence 1.0
source: https://en.wikipedia.org/wiki/Air_gap
tags: [Security, Information assurance]
---

# Air-Gapped Computing

> [!note] Provenance
> Snapshot `air-gapped-computing-000012-dd8534300ecc645e` · content root `dd8534300ecc645e…`
> Derived from <https://en.wikipedia.org/wiki/Air_gap> · Licensed Anticloud Synthetic Licence 1.0

Air-gapped computing is the practice of isolating a computer or network from all untrusted networks, physical and virtual, so that data cannot be exfiltrated across an attack surface.

The assumption of an air gap is that the only paths in or out are physical. In practice the gap has to be maintained deliberately: media must be controlled, maintenance paths must be authorised, and the removal of wireless capability has to be verified rather than assumed from a policy document.

For a knowledge system this is a design constraint rather than a marketing line. If a deployment is to be genuinely sovereign, then the search index, the verification tooling and the integrity ledger must all operate without reaching a third-party service, and that has to be demonstrable on a machine with no route to the internet at all.

That requirement is what rules out mandatory hosted inference, mandatory font delivery from a CDN, and mandatory license validation against a remote server. It is a more productive design constraint than it first appears.
