If the fruit of your labor is yours, so is the record of it. What data sovereignty really means in plain terms, and where the box should actually sit.
Third of the Forge Papers. It takes up one of the five questions the room worked at the Heartland AI Ethics Forum in April: the data off your land, your labor, and your family. Whose is it, and where should it live?
Ask a farmer who owns the grain in his bin, and he will look at you funny. Of course it is his. Ask him who owns the yield map from the combine that filled that bin, and he will stop and think. And if he has read the fine print on his equipment lately, he will get quiet.
That pause is what this whole paper is about.
Locke's argument is simple enough for a child to follow, and old enough that we have stopped noticing it. You own yourself. So you own your work. So you own what your work produces. A free country is built on that chain; every deed, every paycheck, and every contract rests on it.
Now add one more link. The record of your work is produced by your work just as surely as the grain is. The invoice, the field map, the patient chart, the customer list, the ten years of notes on how your shop actually runs. None of that is a leftover that belongs to whoever happens to be holding the hard drive. It is the fruit of your labor in a different form.
That is all "data sovereignty" means. It is not a technical term. It is Locke, applied to a spreadsheet. Your records are yours because you made them.
Here is a piece of history worth knowing. In England, the common land that everyone in a village used was not taken all at once. It was fenced off a little at a time, by people with good lawyers, until one morning the village woke up and found that land it had always worked now belonged to someone else. Legally. With paperwork. And nobody to blame.
The last fifteen years fenced off our records the same way. Not by theft; by terms of service. Every handy new tool arrived with a clause: to use this, your records live on our servers, we get to study them, we may train our systems on them, we may sell what we learn. Nobody forced the farmer to sign. He signed because the tool was useful and the other option was a filing cabinet. And so, a field at a time, the records of American work moved into a handful of buildings in a handful of states, where they are now the most valuable thing those companies own.
Here is the part that should bother you. When one of the big AI systems is trained on "the internet," a good share of what it learned was the working knowledge of small operators who were never asked. Their methods. How they price a job. How they describe a problem. The forgotten man is not in the room, but his labor is in the machine.
Sovereignty is not a feeling. It is a place. Records sit on a physical machine somewhere, and the real question is only whether that machine is yours.
So when I say a business's records should live where the business lives, I mean it plainly. A box, in the building, that the owner can walk over and put his hand on. The AI that reads those records runs on that box, over the shop's own wires. Nothing leaves unless the owner says so. The backups are his. The keys are his. If the software company disappears tomorrow, the records are still in the building, and so is the tool that reads them.
This got practical in the last few years in a way it was not before. There are now capable AI models you can download and run yourself, on hardware a local computer shop already keeps in stock. They are good enough to draft, summarize, and answer questions right on your own machine. When a job truly needs one of the big online systems, you can use it with clear, honest limits on what leaves the building. But the default has flipped. The starting position now is keep it home, and anything that wants to send your records away has to make its case.
There is an old rule, older than this country, sometimes called subsidiarity. The word is fancy but the idea is plain: a job should be done by the smallest, closest group that can do it well, not handed up to a bigger outfit far away. Not because the big outfit is evil, but because it does not know your situation and cannot know it. The knowledge of how your operation really works lives in your people, and it does not survive the trip to a data center two states over.
The economist Friedrich Hayek said the same thing about whole economies. Central planners fail not because they are bad people but because they are missing information: the on-the-ground knowledge that only the person doing the work actually has. A model trained on the average of everybody else's business is a central planner in a nicer suit. A model trained on your operation, sitting in your building, is the man on the ground with a very good memory.
Someone will say, "But the cloud is safer than anything I could run myself." Sometimes that is true, and I will tell you when it is. A big provider has more security engineers than a farm does. But safe and sovereign are two different questions. A bank vault is very safe. It is still the bank's vault.
The real question is not only whether your records are safe from thieves. It is whether they are safe from the people you handed them to. Safe from the terms changing next year. Safe from their systems training on your data. Safe from the company being bought by somebody you never agreed to do business with. Safe from your account being shut off on a Tuesday for a reason nobody has to explain.
A locked door you hold the key to is a different kind of safe than a locked door somebody else holds the key to. Any farmer knows that. So does any parent.
None of this means burning anything down. It means asking three plain questions about every tool you already use.
Property rights did not stop mattering because the property turned invisible. They matter more, because invisible property is easier to walk off with. The records off your land, your work, and your family are yours. They should live where you live. The rest is engineering, and engineering can be done.
The Forge Papers are published one at a time, about monthly, and never to fill a calendar. They are the working convictions behind Liberty Forge: data is property, the human stays in the lead, and efficiency never means a working family loses its living. They are argued in public so a neighbor can weigh them. If one of them changes how you see your own shop, farm, clinic, or classroom, write to me. That is the whole point.