Sovereignty Without Dominance

Rightful authority at human scale, in the age of AI

John Stroh · Director, My Digital Sovereignty Ltd
Research: agenticgovernance.digital

The argument everyone is having

The race, as run, has no winner — and Graylin is right about that.

  • For two years AI has been framed as a race: pick the American stack or the Chinese stack.
  • Alvin Wang Graylin (with Paul Triolo): "there will not and cannot be any long-term winners… on its current path."
  • "adoption beats invention. Distribution beats dominance."
  • His remedy: a "CERN for AI," not a "Manhattan Project for AI."
  • Notice the word he attacks: dominance — not sovereignty.
  • The race is real and intensifying — Iran's data-centre strikes (Mar 2026), the EU sovereignty package (Jun 2026), the Mythos/Fable cut-off. This runs in a different lane: what should be, complementary not contrary.

→ Graylin & Triolo, MIT Technology Review (2025); Beyond Rivalry (2025)

Sovereignty Without Dominance · MDSL · CC BY 4.0

The gaze we have not learned to turn

Scrutinise the US as evenly as we scrutinise China — with respect, not hostility.

  • OECD audiences are trained to mistrust the Chinese state's reach — the National Intelligence Law is real.
  • We are not trained to apply the same lens to the US. Right now the symmetry deserves stating.
  • CLOUD Act (2018): US providers must disclose data "regardless of whether… stored in the U.S. or on foreign soil."
  • FISA §702: warrantless surveillance of non-US persons abroad via US providers (up for renewal April 2026).
  • The EU called the CLOUD Act "a law in possible conflict with the GDPR"; Germany warned off US cloud for sensitive data.
Sovereignty Without Dominance · MDSL · CC BY 4.0

Sovereignty and legitimacy — two measures, not one word

Two axes, not one contested word.

  • Sovereignty (strictly) — control over your own domain, to the exclusion of others.
  • Legitimacy — the rightful authority you hold in the eyes of the people you affect.
  • A continuum the "who's winning" question flattens. Brookings/CEPS (2026): full-stack national sovereignty is "structurally infeasible for almost any country."
  • The race reads only sovereignty-as-capacity at national scale (Huang: "every country needs to own the production of their own intelligence") — which most will never hold.
  • What most of the world can hold: legitimacy + a bounded data sovereignty.

Brookings/CEPS, Is AI Sovereignty Possible? (2026) · Huang/NVIDIA, the national pole

Sovereignty Without Dominance · MDSL · CC BY 4.0

What the Village holds

The sovereignty worth wanting is a community's control over the data and systems that act on it, exercised with legitimacy — held at human scale, federated, never surrendered.

A coordinate on the map — high legitimacy, modestly but genuinely sovereign over its own domain. Two things, co-equal, as co-governance:

  • Legitimacy — the community's rightful standing to decide what its AI may do, must refuse, and who answers for it.
  • Data sovereignty — control over the one domain it can hold: its data and the models that act on it (rangatiratanga over data).

Not dominance over anyone — legitimacy plus the capacity to say no. It scales by federating outward, not racing upward. A federated mesh has no node that must win.

Sovereignty Without Dominance · MDSL · CC BY 4.0

This is old, not invented

Four traditions, no shared lineage, one conclusion.

  • Elinor Ostrom (Nobel, 2009): communities can self-govern commons — a third way past privatise-or-nationalise; eight design principles, including nested enterprises (federation).
  • Subsidiarity: authority at the lowest competent level — Quadragesimo Anno §79 (1931); EU law.
  • Simone Weil: obligations come before rights; rootedness is a basic human need.
  • Te ao Māori: rangatiratanga — relational authority; Māori data sovereignty grounded in Te Tiriti.
Sovereignty Without Dominance · MDSL · CC BY 4.0

The architecture — an immutable constitutional layer

"Never surrendered" must be a property of the system, not a promise.

  1. Immutable floor — inviolable values enforced in architecture, not policy: human-in-the-loop on certain decisions; restricted content forces refusal/escalation; te Tiriti as a constitutional commitment, not a clause. No single party — operator, community, or model — can override or stealth-erode it.
  2. Community rules above the floor — governance precedes infrastructure.
  3. Co-equal authorities — each publishes rules and can withdraw them at any moment.

The immutability is the structural answer to values-creep: values that can't be quietly overridden can't be quietly normalised away.

Sovereignty Without Dominance · MDSL · CC BY 4.0

Why now — normalisation, and a closing window

Arendt: great harm arrives banal — by "a curious, quite authentic inability to think." Extractive AI normalises by habituation, not decision; guardrails erode by familiarity. The immutable floor keeps core values out of reach of that drift.

But the timing cuts both ways:

  • The same exponential capability that speeds the drift makes the alternative cheap to build.
  • The research behind this essay was done by a swarm — 100+ sub-agents in one run, minutes not days, twice over.
  • A small, principled team can now stand up a sovereign, multi-region system on situated small models (Llama, Qwen), outside any single state's reach, in months.
Sovereignty Without Dominance · MDSL · CC BY 4.0

A live exhibit — and an honest admission

It happened while this was being written, at my own expense.

  • 12 June 2026: a US export-control directive suspended Fable 5 + Mythos 5 for "any foreign national, whether inside or outside the United States."
  • The most capable model — withdrawn overnight from every foreign national, by an order they had no part in.
  • Anthropic disagreed publicly and is working to restore access. The company is not the villain. There is no villain. There is only the structure.
  • Admission: this essay was drafted with US models on US infrastructure — the tool arguing for a sovereign stack is itself subject to the dominance logic. The argument, demonstrated — not hidden.
Sovereignty Without Dominance · MDSL · CC BY 4.0

The aspiration — and the ask of the reader

Every community able to run AI it governs, and own the data it creates — honouring te Tiriti and other co-equal principles in an immutable constitutional layer no one can quietly override.

  • Not every community winning the capacity race — almost none will.
  • Every community holding the authority it is entitled to — at human scale, federated, never surrendered.
  • A quiet act of non-alignment: owing allegiance to no patron, because it doesn't play the patrons' game.

Before asking whose AI to build on, ask what authority you are entitled to hold yourself — and then hold it.

Sovereignty Without Dominance · MDSL · CC BY 4.0

An essay about a single word. Sovereignty has been conscripted to mean the capacity to dominate a technology. There is an older, better meaning: rightful authority over the systems that act on you — at community scale, federated, never surrendered. Respect for every party, enmity toward none, steadfast on one principle.

The Village begins by agreeing with all of it. But agreeing the race is a dead end is not denying it is real — it is an accurate description of what is emerging; this runs in the normative lane, complementary not contrary (a two-lanes framing owed to Tim Clancy). The essay extends Graylin in two directions his frame leaves open: the legitimacy/sovereignty distinction, and even-handed scrutiny.

The US is not China — differences in transparency, courts, and recourse matter, and the essay says so plainly. The narrow, sturdy point: for a community that is neither American nor Chinese, dependence on EITHER stack is a standing exposure to a decision it has no part in. The reflex to scrutinise only one side is a blind spot — where authority quietly leaks away.

Most actors will never hold capacity-sovereignty. A "CERN for AI" convenes the powers large enough to be invited. The authority of the governed — a community's standing over the systems that act on its own people — can be held today by actors who hold no frontier cluster.

Authority without data self-determination is a seat at a table where someone else holds the records. Data self-determination without authority is a vault with no standing. Together they describe a sovereignty available to most of the world.

Federation is not a pattern we invented — it is what enduring commons look like. Te ao Māori is named with care, as one tradition among several, never spoken for; Te Tiriti carries real NZ legal weight, the wider instruments do not, and that is not overstated. The convergence across unrelated traditions is the reason to trust the principle.

A community may make the floor stricter for itself; none can be forced below it. Legitimacy is composed from co-equal parts, never concentrated.

Arendt held the individual fully responsible — the banality is in the mechanism, not the culpability. The capability that accelerates the danger is the same one that finally arms the response. There is no longer an excuse to wait, and not much time in which waiting is safe.

A capable practitioner in NZ lost his most powerful tool to a sovereign he does not answer to. That is sovereignty-as-dominance reaching down through the layers to a single person — the dependency the Village posture declines. One governs the layer one can hold, and is honest about the layer one cannot yet.

An ambition held firmly, not a boast. The Village does not claim to have solved AI governance or to rival a frontier lab. It claims the layer where rightful authority can be held is buildable now — for actors who will never hold the other kind. Respectful of every interest, hostile to none, steadfast on the one principle worth being steadfast about.