How this was made. The version number counts drafts of the text. It does not measure the inquiry behind it, which has run over days and across several AI systems, with argument between those systems and within them, directed, refused and repeatedly redirected by the author. The source material was AI-generated, and then adversarially and iteratively refined across a range of tools — systems built by different companies in different jurisdictions, set against each other and against the author. No one of them produced this text, and no one of them reviewed it alone. The plurality is deliberate rather than incidental. A single model carries a single set of priors about which sources are authoritative, and this series argues that an evidence base narrowed in exactly that way is how a contested question comes to look settled. Using one model to investigate that claim would have been the claim refuting itself. To name a single model on it would credit that model with work that was neither its own nor done in a single pass. The plurality was also necessary, and the record should say why. In drafting, the assisting model repeatedly led with United States institutional sources — a national laboratory, an industry association, a market study nineteen years old — and presented conclusions drawn from them as the state of knowledge. On one occasion European measured data contradicting those conclusions was present in the same research return and was placed below them. Framings were proposed that would have argued against this series’ own position using that evidence base, and offered as rigour. Each was refused by the author and the material rebuilt. That is the mechanism these documents describe, occurring in their own making, and it is recorded because a series arguing that evidence bases narrow without anyone deciding to narrow them cannot credibly claim its own production was exempt. The framing, the corrections and the judgements are the author’s, and so are the errors. How this site is written sets out what is declared on every piece, who checks it, and where the per-piece record lives.
The five questions are the same at any size. What changes is the length of the answer.
A three-person practice can answer “who can stop this” in one sentence and be right. A government department may need a page and still be wrong. The small organisation’s disadvantage is that it will usually get no useful answer from a supplier — and the piece is direct about that: a club or a practice asking these questions of a large vendor will generally be told what the vendor is willing to say.
That is worth knowing before signing rather than after.
It is one of the four controls and not the other three. Owning the data does not tell you whether you can inspect what the system did with it, change what it is permitted to do, or stop it.
The EU’s Data Act, at Article 23, obliges the removal of obstacles to switching and porting, which is a real lever on the ownership question. It is not a lever on the other three, and an organisation that reads its data clause and stops there has answered one question out of four.
Their definitions are worth reading closely, and the piece quotes three. NVIDIA’s is “a nation’s domestic capability to produce artificial intelligence using its own infrastructure, data, workforce, and business networks.”
Read the terms in it: infrastructure, data, workforce, networks. Now look for a term covering inspection, alteration, or stopping. There is none, in any of the three examined. That is not an accusation of bad faith — the definitions are about production capacity, and they are coherent definitions of production capacity. They are simply not answers to the question of who holds the controls, and they are quoted as though they were.
On 10 June 2025, a director of Microsoft France was asked before a French Senate commission of inquiry whether he could guarantee that data would not be transmitted to a foreign authority without the agreement of the French state. His answer, on the record: “Non, je ne peux pas le garantir, mais, encore une fois, cela ne s’est encore jamais produit.”
He was being straightforward. The point is not that he refused a guarantee; it is that the guarantee was not his to give, and no contractual language changes which jurisdiction a parent company sits in.
Te Kāhui Raraunga published the Māori Data Governance Model in 2023 — Tuia te korowai o Hine-Raraunga — with eight pou, authored by Kukutai, Campbell-Kamariera, Mead, Mikaere, Moses, Whitehead and Cormack. It sets out governance of data by the people to whom the data relates, and it predates most of the vendor material on sovereignty.
It is cited because it is a developed body of thinking on precisely this question, not as a borrowing of authority. Its standing is a matter for the bodies that hold it.
An organisation that cannot answer any of them and is nonetheless able to set a limit that holds, find out what happened, change the arrangement, and stop it — repeatedly, and not by the supplier’s goodwill on the day.
If that turns out to be common, the five questions are measuring something that does not matter and should be replaced.
What this publication does not claim, and what is outstanding against it in the register.
Nothing outstanding in the register. Every claim in this publication has its evidence recorded, and no question against it is parked. That is a statement about this publication on the date shown above, generated from the register rather than asserted, and it will change when the register does.
Alongside: the publication · sources and provenance