Whose data, whose rules

The first thing a farm does with a new tool is agree to it. Not use it. Agree to it. Somewhere between the corn and the cows, somebody sits at a kitchen table with a laptop, or thumbs a phone in the cab of a tractor, and up comes a box of text that nobody on the farm wrote. The text is not about how to make the tool work. It is about what the tool is allowed to do with everything the farm tells it from that day onward.

A person can read that box or not. Most of the time, not. The company that wrote it is counting on that, and the company is usually right. The sentence people reach for, when they do read it and get as far as feeling uneasy, is the wrong one. Whose data is it?

That question has an answer, and the answer does not help. A farm can own every byte of its own record and still hold almost nothing. One field of yield maps, one herd of milk numbers, one season of sprayer logs: on their own they are worth about what they weigh. Data is one of the only things on a farm that is worth more the more it is combined. A thousand yield maps across a dry year become a map of what the season actually did. A thousand herds of records become a picture of a whole region's health. The value is in the joining, and the joining never happens on one farm.

So the useful question is not whether the data is yours. Yours and alone is a weaker place to stand than pooled and read. The useful question is the one the box of text was written to answer in advance: when the records combine, who writes the rules the combination runs under, and who gets to read what comes back?

That is not a property question. It is a rules question. And the people who get a say in the rules are the people who can read them.

The words that make people nervous

Jennifer MacTavish, interim executive director of the Agricultural Adaptation Council, sat on a Future Herd panel in June 2026 and put her finger on exactly this. Ask a farmer about artificial intelligence and you get unease. "when I talk to farmers about ai, they get nervous," she said. Change one word and the same person gives you a different conversation. Talk to the same farmer about data and decision making, and they "know exactly what they need, what they're nervous about, and the kinds of data they need to make decisions."

The nervousness is not about the machine. It is about being asked a question nobody explained, in a language the field does not use. MacTavish said it plainly: "it's just the words we're using that are making people nervous."

The farmers she describes are not confused about their farms. They know their ground, their animals, their numbers. What they are handed is a vocabulary that does not belong to the field, wrapped around decisions that do belong to it. The person who can sit between those two, read the box, and say what it actually means for this one farm is not a lawyer and not a data scientist. They are a reader.

What "sovereign" would mean

Mohamad Yaghi, a vice president at Farm Credit Canada's innovation hub, was on that same panel and turned reading into a concrete test. The return on a tool, he argued, is not only profit. It is also whether a farm can "manage that data on premise within Canada and keep our data sovereign as well," rather than ship the record out by default.

Sovereign is one of those words that gets waved around, so here is what it actually means. Sovereignty over data is not a claim about who holds the file. It is a claim about whose rules the file moves under. A record stored in Canada but governed on a California server under a California terms of service has not stayed in Canada in any way that matters. Yaghi put the harder form of the question to the equipment makers, since they are "some of the biggest providers of that data, but I don't know how much of that information is within Canadian soil."

That is the question, and it has nothing to do with ownership. It is about where the rules are written, and whether anyone the data belongs to had a hand in writing them.

The rules are written in the terms

The vendor side answers the ownership question without flinching. Brian Leake, a spokesperson for Bayer, put it in 2026 in the plainest form it takes: "Farmers own their own data — full stop."

And here is the same arrangement from the seat of the tractor. Andrew Nelson is a fifth-generation farmer in eastern Washington who runs several thousand acres and uses both a chemical-recommendation platform and an equipment maker's operations centre. He has read the agreement. "I've read the terms and conditions." What he cannot square is the other half of it: "But giving some of that information to the person that I'm having to buy a three-quarter of a million-dollar machine from just doesn't sit quite right."

Both sentences are true, because they are about two different things. "Farmers own their own data" is a sentence about a file. The terms are the sentences about the rules. A whole certification industry has grown up around making those rules visible: a vendor can be audited on whether it discloses its data terms in a standard form, and the audit does not limit what the terms may say. A contract can pass that check and still let the vendor fold a farm's records into an anonymized pool it controls. Visibility is not the same as having a say.

So the ownership sentence and the operating rules travel in the same document, and the two do not contradict because they barely meet. The contradiction only appears when a farmer asks the rules question instead: what happens to the pooled version, and who decided?

There is no law behind this

Most people assume there is a law here. In Canada there is not one that reaches the thing this chapter is about. The federal privacy law covers information about identifiable people, and a yield map is not about a person, a seeding rate is not about a person, a milk component is not about a person. So the statute most Canadians are thinking of when they say privacy does not touch farm operating data.

What reaches it instead is nothing, and researchers have named the gap precisely. Two Canadian scholars, Sarah-Louise Ruder and Kelly Bronson, wrote in Policy Options in 2024 that Canada does not even have a legal definition of agricultural data. Before a rule can govern a thing, the thing has to be a category a rule can point at, and the records this chapter is about have no such name in Canadian law.

Reform has been attempted and has stalled, and the bill now before the House is about people's information, not the farm's operating records. Which leaves the rules where they have always lived: in private terms, agreed one click at a time, in documents almost nobody reads. That is why the terms matter as much as they do. They are not the fine print under the real thing. In farm data, the fine print is the real thing.

The good version: rules written by the people whose data it is

None of this means the rules have to be written against the farm. The good version is ordinary, and it has been running for decades.

On a dairy farm on milk recording, a technician comes by about once a month. Milk from each cow goes into a sample bottle, the bottles go to a laboratory, and the numbers come back: what each animal gave, how much butterfat, how much protein, how many cells should not have been there. That report is not paperwork. It is the herd's memory with units attached, and it is how a farm decides which cow gets bred back and which one has run out of chances.

The telling part is who holds it. In Canada the national record of the dairy herd is held by Lactanet, a farmer-run organization formed in 2019 when three regional herd-improvement bodies merged, paid for and owned by the farms it serves. The farms are not customers of it. They are the reason it exists. And when the international body that decides how animal records are shaped writes its next version, the Canadian farmer-run organization sits on that working group, helping write the rules its members will farm under.

The reason this matters is not that the data stays on the farm. Milk records have to travel: a veterinarian needs them, a nutritionist needs them, a processor needs some of them. The difference is who decides how they travel, and who can read the whole picture.

Donald Killorn, executive director of the Prince Edward Island Federation of Agriculture, described the goal that a farmer-run arrangement reaches. Work with the equipment makers, yes, without being locked inside their platforms, so that manufacturers can hand records to each other and "farmers will have a great deal of control over how that data is collected and used." And he drew his own line on where the record lives: "we will not be moving all of the data off farm."

He also said what the rules are actually worth, which is the sentence that turns this from a technical matter into a business one. Data governance, he told the panel, is becoming "the most important thing to profitability and business risk management." Not a chore you comply with. The thing that decides whether a farm makes money and whether it survives a bad year.

Suresh Neethirajan, who works on digital livestock systems at Dalhousie University, asked the ownership question in exactly the terms this chapter is setting aside, and then answered it the way this chapter does. "Hey, who owns the data?" The easy half of the answer, he said, is that "the data belongs to the farmer and of course the animals themselves." The rest is harder, and it is the part that matters: the question actually lands in this country with Dairy Farmers of Canada, the marketing boards, and the co-operative societies. Ownership is a sentence anyone can say. The rules are made in the institutions farmers already run.

The farmer as hacker

There is a name for the person this subject keeps turning into, and it is a better one than data literate. The farm has always had them: somebody who reads a system to understand it, who takes something apart to see how it works, who makes a tool serve the work rather than the other way around. That person has an old name, and the name is hacker.

The word has been dragged through decades of headlines into meaning the person who breaks in. That was never the point. A hacker, in the sense the word carried when it was new, is the person who refuses to be a passive user of a machine and reads it instead: learns how it works, finds where it could do more, changes it. The farm has been doing this since before the word existed. Which cow is off her feed, which gate sticks when the ground freezes, which field floods first. Nobody handed the farm a manual for any of those, and the farm did not need one. Reading the system and improvising on it is the oldest job on the property.

That is why the answer to "whose data is it" is not to get scared and hoard. A farm that guards its record like buried treasure makes itself blind, because the record is only worth anything in company. The hacker's answer is different. Read the terms. Learn what the tool actually does with the record. Find the lever that moves it, and the gear that would have to change for it to move further. Figure out which rules you can bend yourself, and which institution you would have to own or join to change the rest. Then bend the tool back toward the farm.

That is a role, not a chore. It is the difference between being a customer of a platform and being a maker of the arrangements. The farmers MacTavish described, the ones who know exactly what they need as soon as the words stop being jargon, are already doing this. They have spent generations reading living systems. Data is just a new alphabet for the same skill.

Elinor Ostrom spent a career asking why some things held in common stay healthy while others collapse, and her answer is the sentence this chapter has been circling. A commons lasts when the people who bear the cost of its rules are the ones who write them. Boundaries, the means to settle disputes, the power to change the rules as conditions change: those are the load-bearing walls, and ownership barely enters it. A pool of farm records is a commons, and a commons runs on its rules.

The caution travels with the hope. The co-operative form is not the answer by itself, any more than open is; a data co-op with no maintenance plan fails the way an unmaintained open-source project fails, only more expensively, because it takes a community's trust down with it. And at the far edge of the argument, pooling has been worn as a brand by a venture-backed company that raised close to a billion dollars from Google's venture arm and a Singapore state fund on the strength of other people's records. Combining data is only good when the people who combine it write the rules, and earning the right to write them is the work.

A line this chapter will hold

There is one boundary this chapter will not cross, and it is the sharpest form of the point it has been making.

Some records should not be combined, published, or shared at all, no matter how carefully the rules are written, because they belong to the people they are about, and those people decide. In Canada, the First Nations principles known as OCAP, for ownership, control, access and possession, were established in 1998 and are stewarded by the First Nations Information Governance Centre. Internationally, the CARE principles, for collective benefit, authority to control, responsibility and ethics, were written as a deliberate companion to the open-data principles, because openness alone said nothing about who holds the knowledge being opened.

This book is not the place to summarize those frameworks. A farmer in eastern Ontario has no standing to restate a governance system built by and for peoples whose relationship to land is not mine to describe. The frameworks are named here and pointed at directly because they settle the question this chapter keeps asking: the first question is not what the data says, and not who owns the file, but who governs, and who decided.

That cuts against this book's own instinct, and the cut stays. Every chapter here argues that people should be able to inspect, run, modify and repair the systems they depend on. Followed with no conditions, that argument points toward publishing everything. OCAP and CARE say plainly that some things should not be published, and that the community the knowledge belongs to decides which is which. If the book's case for openness cannot survive that, it was never a case for openness. It was a case for everyone getting to see everything, which is a different and worse idea. Open is not automatically the good side. The work is the governance, not the file.

Open questions

Does a farm need to own the file, or does it need a seat at the table where the rules are written? The co-operative record so far says those are not the same thing, and that the seat matters more.

What would make the joined picture come back to the farms that built it, so that a thousand yield maps read back as a region's memory instead of a supplier's negotiating position?

If Canada ever wrote a law for farm data, what would it attach to, when nobody has agreed on a definition of the thing?

If the farmer is the hacker, and the hacker is the one who reads the rules and bends the tool, then what is the first tool a farm should learn to read, and who teaches it?

Moves you can make

Pick one tool you already depend on, and find its data terms. Not the marketing page. The terms. Read the section on what the company may do with data that has been aggregated or anonymized, and the section on what happens to your data if you stop paying. Keep a dated copy, because terms change, and the version you agreed to is the one that governs you.

Then write one sentence, in your own words, saying what that document gives the company and what it gives you. One sentence is enough. If you cannot write it, that is the finding, and it is worth more than a summary somebody else wrote.

Then do the hacker thing. Find out whether the tool lets you take your record with you. Not whether the company says it will export. Walk it: find the export button, make a copy, open it, and see what is actually in it. Most tools answer this question differently in the advert than in the file.

The situation: one policy you have already agreed to

Think of one thing you use that holds information about you or your work: the barn software, the field app, the grocery loyalty card, the fitness tracker, the ride home. You agreed to its terms. You almost certainly did not read them. Find them. Read the data section all the way through, and read it as the person on the other side of it, because you are. Then rewrite one clause in plain language, the way you would explain it to somebody at your kitchen table, and keep both versions.

Two things will happen. The first is that it will take longer than you expect, because the sentences are built not to be read. The second is that you will find at least one clause that a reasonable person would have refused if it had been said out loud.

Now take one more step, the one that separates a reader from a hacker. Find out who else is bound by the same terms. Your neighbour, your co-op, the six other farms on the same platform. A clause that is unacceptable to one farm is a complaint. A clause that is unacceptable to enough farms to matter is a negotiation. Somebody wrote the rules you just read. Somebody who reads them well enough gets to write the next ones.

The system the farm is inside

Every chapter so far has been about a system the farm uses. The next one is about the system the farm is inside, and the one thing on this farm nobody gets to negotiate with: weather that no longer keeps its old pattern. The record this chapter has been arguing about has a second use, and it is the one that matters most. A farm that can read its own history holds the only thing adaptation actually runs on: a memory of what this particular ground did the last time it was this dry.