01Neev AI
02Swar
03Synvio04Terrasyn05NetOFF06Proppel07About08Contact
EvioraxNetOFF
Air-gapped AI · Offline LLM

Your own GPT.
Offline by design.

NetOFF is a private large language model that runs entirely on your own server or GPU. Your prompts and documents never travel to ChatGPT, Gemini, Claude, or anyone else's cloud.

Book a demo See the air gap
NetOFF · Node 01Uplink: none
You

Summarise this tender file and flag every clause that needs legal review.

NetOFF

Summary ready. Three clauses need a lawyer before you sign:

  • Penalty terms — clause 7.2
  • Data-sharing obligations — clause 11
  • Arbitration venue — clause 14
Outbound traffic
0 bytes
Cloud API calls
0
Inference
Local GPU
Illustrative session

That cable is the internet. Try plugging it in.

The air gap

Follow a prompt. See where it goes.

An air gap is the oldest rule in security: a machine with no connection to the outside cannot be reached from the outside. Send a prompt down both routes and watch which one leaves your building.

Route traceOne prompt · two routes
Through a cloud AI assistantReady
  1. Your deskYou
  2. Office networkYour IT team
  3. Your perimeter
  4. Internet providerA telecom company
  5. Public internetMany networks
  6. AI provider's data centreThe AI company
Through NetOFFReady
  1. Your deskYou
  2. Your NetOFF serverYour IT team
  3. Your perimeter
  4. Nothing goes past this line.

Pull the network cable out of the wall and NetOFF keeps answering. It never needed the cable in the first place.

What it does

Ask it something. In private.

The same everyday help people already reach for, without the part where the question leaves the building. Pick a task and watch the counter.

Local sessionAsk & explainNetwork off
You asked: Explain this circular in simple terms. NetOFF answered: In short: from next month, any purchase above the approval limit needs two sign-offs instead of one. Who it affects: anyone raising a purchase request. What changes: a second approver is added to the form. What to do now: nothing, until the new form is issued.
Outbound traffic
0 bytes
Cloud API calls
0
Answered on
Local GPU

These answers are scripted examples. NetOFF runs on your hardware, not on this website, so the real thing can only be shown on a machine. Book a demo to try it on your own files.

Cloud AI · NetOFF

Same questions. Different destination.

Cloud assistants are excellent tools for work that is allowed to leave the building. NetOFF is for the work that isn't. Ask the questions your security team will.

Cloud-hosted AI

Over the internet to the provider's data centre.

NetOFF

Nowhere. It is processed on the machine in your building.

ChatGPT, Gemini, and Claude are trademarks of their respective owners and are named on this page only as examples of cloud-hosted AI services. No affiliation or endorsement is implied. The comparison describes how cloud-hosted AI works in general, not any one provider's terms.

Government · Private sector

Built for work that can't go to the cloud.

If a document would never be emailed to a stranger, it shouldn't be pasted into a public chatbot either. Pick your sector to see what stays inside.

Your sector
Stays inside

Citizen records and internal files

What they'd ask NetOFF

Draft a reply to this RTI application.

Data residency by constructionNo vendor in the data pathYour audit scope, unchanged

Deployment

Your hardware. Your call.

NetOFF is sized to the machines you own and the people who will use it. Answer two questions to see the shape yours would take.

Who will use it?
How isolated does it need to be?
A setup like yours

On-premise GPU server

Fully air-gapped

A GPU server in your own server room, reached from any office browser over your internal network. No external link at all. Updates come in offline, on your terms, never over a live connection.

Scale
Size it with us
How it goes in

From first call to sealed room.

NetOFF is installed, not signed up for. Step through what that looks like, start to finish.

Scope

We map the use cases, the people who will use it, and how sensitive the material is.

Together, in a first call
From the Eviorax lexicon
Entry 05 · 06

NetOFF

/ˈnɛt·ɒf/ · noun

Coined — English net + off — “the network, switched off”

Eviorax productA private large language model that runs on your own server or GPU — a GPT-style assistant that keeps answering with the network cable unplugged.

Built for
Government departments, defence, banks, hospitals & enterprises
Runs on
Your server or your GPU — on-premise or fully air-gapped
Connects to
Nothing outside your perimeter
Pairs with
Swar — voice AI that runs inside the same perimeter
Built in Bharat · Ready for the world

Every country has data that cannot leave.

Every government and every regulated industry on earth holds information that may not cross a border or touch a third party. NetOFF doesn't negotiate with that rule — it runs on the inside of it.

Ready forAny jurisdiction

Residency by construction

Nothing leaves the machine, so there is no cross-border transfer to assess. The data-residency question is answered by architecture, not by a clause in a contract.

Architecture, not adaptationBuilt in Ahmedabad, India
One build · every market
Questions worth asking

Things worth asking.

Security teams ask hard questions, and they should. If yours isn't here, we'd rather answer it on a call than leave it to a datasheet.

Talk to us
What is NetOFF?
NetOFF is a private large language model that runs on your own server or GPU. It works like the GPT-style assistants people already know — it answers questions, drafts, summarises, and translates — but the model lives inside your organisation instead of on someone else's cloud.
Does it really work without the internet?
Yes. The model and everything it needs sit on your hardware, so it answers with no internet connection at all. For the most sensitive environments it runs fully air-gapped, on machines that have no external link.
Is any of our data sent to ChatGPT, Gemini, Claude, or any other cloud AI?
No. NetOFF does not call any external AI service. Prompts, uploaded documents, and answers are processed on your hardware and stay there. In an air-gapped deployment there is no route out even in principle.
What hardware do we need?
It depends on how many people will use it and what they will ask of it. A single GPU workstation can serve a small team; a department runs on a GPU server. We size the deployment with you before anything is ordered, and start from hardware you already own where we can.
Can it answer questions from our own documents?
Yes. Policies, manuals, case files, and SOPs can be indexed locally so staff can ask questions of your own material. The index is built and stored on your hardware, like everything else.
Is it as capable as the big cloud assistants?
The largest cloud models run on data centres no single organisation owns, so on raw breadth they lead. NetOFF is built for work where sending the data out is not an option — drafting, summarising, translating, and answering from your own files. The honest test is your own tasks, which is what we run in a demo.
How does it get updates if it is offline?
Updates are brought in offline, on your schedule, and under your own change-control process. Nothing is pulled over a live connection, and nothing is installed that you have not approved.
How do our staff use it?
Through a chat window in the browser, on your internal network. There is nothing to install on each desk, and anyone who has used a chatbot will recognise it immediately.
What does a pilot look like?
One team, one machine, your real work. We set NetOFF up on your premises, your staff use it on their actual tasks, and you check for yourself that nothing leaves the building before deciding whether to scale.

Go on, try switching it on.

Bring the AI inside.
Leave the internet out.

Start with one team and one machine. See what a private assistant does with your real work — and check for yourself that nothing leaves.

Book a demo Back to Eviorax