The Aitronos Box

Some of your data is never going to the cloud.

That rule is not the problem — the problem is what it protects. The records that would make the biggest difference are the exact ones a regulation, a contract or a policy keeps away from AI.

The Aitronos Box is the whole platform, on a machine in your building. The rule stays. The work happens anyway.

Outside Model provider Cloud provider Sub-processor Stopped at the wall Your building The questions people ask AITRONOS BOX Plant room The question The answer Outside Model provider Cloud provider Sub-processor Stopped at the wall Your building The questions people ask AITRONOS BOX Plant room The question The answer
The boundary — nothing on the left ever reaches the right

What it is

It is a machine, and it goes in a room you control.

We specify it, we buy it, and it arrives ready for your network. The whole platform, not a cut-down edition.

Inside the machine · what actually runs on it 01 / 03
Outside The public internet Model vendors Aitronos Nothing crosses this line A room you control The Aitronos Box The models The knowledge The automation The people asking Upstairs, at their desks Your systems Already in this building Your documents On your file shares The people asking Upstairs, at their desks The models On your GPUs Your documents On your file shares The knowledge Your storage Your systems Already in this building The automation Flow-Plate

The models. Generation runs on the GPUs in that machine. A prompt carrying a contract or a patient record never leaves it.

The knowledge. Your documents, and the index built from them, sit on your storage. The passage that answers a question never travels.

The automation. Flow-Plate, our workflow engine, runs where your systems already are and reaches them directly — not through an opening in your firewall.

You can walk down the stairs and put your hand on it.

Case files Patient records Contracts · NDAs Salary data Board papers Twenty years of research Not for AI
The drawer, and what is in it

The payoff

The drawer you keep shut.

Every company has one, and everyone knows what is in it: the most valuable material in the building, and the one place AI has never been allowed to go.

A Box is what opens that drawer. Ask questions of it. Draft against it. Point workflows at it. There is no sub-processor list, because there is nobody else on it.

Nobody can be compelled to hand over what they do not hold. On a Box, we do not hold it.

The question that stalls every review — where does our data go? — stops being a diagram with arrows leaving the page. It becomes a room number.B·04

Before you call

Three things you are already thinking.

“We would be running AI infrastructure.”

We specify the machine and we buy it. Who operates it, how it is updated and what it costs are settled on the scoping call.

“Which models run on it?”

Open-weight models, on your GPUs. The largest commercial models cannot leave their providers' data centres, and no machine you can buy changes that.

Frontier model, in the cloudNot allowed near it
Open-weight, on your GPUsRuns
What happens todayNothing

The comparison that decides it

Be careful what you compare it to. For the drawer that stayed shut, the alternative was never a bigger model. It was nothing.

“What about permissions, the record, and getting our data out?”

Unchanged. Who can reach what, what is written down and what you can export work the same either way.

Read Trust & Platform — permissions, the record, GDPR by article →

Customised hardware

Your Box is specified, not ordered.

No two are the same — the machine is sized to the models, the documents and the systems it has to serve. Bring the rule you cannot change, and you leave the first call knowing what a Box for it looks like and what it costs.

Swiss-hosted, or hosted by you. If the Swiss-hosted platform already covers your rule, we will tell you on the call rather than sell you hardware.

AITRONOS BOX AITRONOS BOX YOUR UNIT YOUR UNIT ASSEMBLED FOR ASSEMBLED FOR
Built to order, engraved for one owner