# What Is Air-Gapped AI? A Plain-English Definition

> Air-gapped AI runs on a system that has no physical connection to other networks; data moves in or out only by hand, under human control. The model, updates and licences therefore have to be carried in through a controlled transfer process.

Canonical URL: https://www.ruebstudio.de/en/glossary/air-gapped-ai  
Author: Dr.-Ing. Marcus Rüb, CEO  
Published: 2026-11-05

Air-gapped AI is an AI system that runs on hardware with no physical connection to any other network, so data enters or leaves only by hand. The [NIST glossary](https://csrc.nist.gov/glossary/term/air_gap), citing CNSSI 4009-2022, defines an air gap as an interface where systems are not physically connected and any logical connection is not automated. That is a stricter idea than [offline AI](https://www.ruebstudio.de/en/glossary/offline-ai), which only needs to work without a connection.

## How does an air-gapped LLM work in practice?

At runtime, an air-gapped LLM behaves like any [local AI](https://www.ruebstudio.de/en/glossary/local-ai) setup: weights, inference server and data sit on your own machines. What changes is the supply chain. Everything the system needs, from model files to patches, has to arrive through a controlled, manual transfer.

Think of it as three supply lines:

- **Models.** Download on a connected machine, carry across on approved media, drop into the runtime's model directory. Ollama reads models from the folder set in `OLLAMA_MODELS` ([Ollama FAQ](https://docs.ollama.com/faq)). NVIDIA documents a fixed routine for NIM: fetch a profile with `download-to-cache` on a connected host, copy the cache, then start the container on the isolated host without an `NGC_API_KEY` ([NVIDIA NIM air gap deployment](https://docs.nvidia.com/nim/vision-language-models/latest/deploy-air-gap.html)).
- **Software.** `pip download` saves packages to a directory that you can later pass to `pip install --find-links` ([pip documentation](https://pip.pypa.io/en/stable/cli/pip_download/)). Container images and OS packages need an internal mirror of their own.
- **Licences.** Anything that phones home to validate a licence or needs an online account will stop working. Ask vendors about offline activation before you buy, not after.

Tools also need to be told not to reach out. With `HF_HUB_OFFLINE=1`, the Hugging Face library makes no HTTP calls and serves only cached files ([Hugging Face docs](https://huggingface.co/docs/huggingface_hub/package_reference/environment_variables)). Ollama has a local-only mode via `OLLAMA_NO_CLOUD=1` ([Ollama FAQ](https://docs.ollama.com/faq)).

## Air-gapped vs on-premise vs offline

These terms describe increasing levels of isolation. An [on-premise LLM](https://www.ruebstudio.de/en/glossary/on-premise-llm) runs in your own data centre but usually sits on the corporate network, often with internet access. Offline AI does not need a connection while running. Air-gapped rules out a connection physically and governs every transfer with a procedure.

So "air-gapped" describes how you operate a system, not a product feature. The same installation can be connected during testing and isolated in production.

## What to watch out for

The transfer path is the weak spot. A USB drive that carries a model can carry malware just as easily. Typical safeguards are approved media, checksums for every file, a scanning station before import, and a log of who moved what and when.

There are trade-offs in daily use too. Agents cannot call web search, external APIs or hosted models, and security fixes arrive only as fast as your transfer process runs. Air-gapping is worth it when data must provably never leave the network, as in defence, critical infrastructure or regulated research. For many mid-sized companies, a local deployment behind a strict firewall, as outlined under [sovereign AI](https://www.ruebstudio.de/en/sovereign-ai), is enough. For running models without internet in general, see [AI without internet](https://www.ruebstudio.de/en/guides/local-ai/ai-without-internet-offline).

## Frequently asked questions

### Is a system behind a strict firewall air-gapped?

No. A firewall filters a connection that exists. An air gap means there is no physical link at all, and every transfer is manual.

### Can I use hosted models such as ChatGPT in an air-gapped setup?

No. Only models whose weights you can download and run entirely on your own hardware will work.

### How often should you update an air-gapped AI system?

On a fixed schedule that the transfer process can handle, plus ad hoc for critical security fixes. The cadence is an operational decision, not a technical limit.

## Sources

- [NIST CSRC Glossary: air gap (from CNSSI 4009-2022)](https://csrc.nist.gov/glossary/term/air_gap)
- [Hugging Face Hub: Environment variables (HF_HUB_OFFLINE)](https://huggingface.co/docs/huggingface_hub/package_reference/environment_variables)
- [Ollama FAQ](https://docs.ollama.com/faq)
- [NVIDIA NIM for VLMs: Air Gap Deployment](https://docs.nvidia.com/nim/vision-language-models/latest/deploy-air-gap.html)
- [pip documentation: pip download](https://pip.pypa.io/en/stable/cli/pip_download/)
