Class: where the framework may be strong — strong normative diagnosis, weaker transition mechanism. It runs on the hard standard below.
The threat
If a handful of firms or individuals own the models, the compute, the data, the robotics, and the wealth AI generates, the returns concentrate faster than any previous technology allowed, because the marginal cost of another unit of “labor” collapses toward zero for whoever owns the machines. But the deeper danger is not that some people become extremely rich. It is that the entities controlling AI acquire the practical ability to decide who gets access to advanced capability, which work remains viable, which institutions survive, what knowledge is available, what infrastructure people depend on, which communities can build alternatives, and eventually which forms of participation remain realistic. At that point ownership concentration becomes ecological concentration: the field may still look abundant — extraordinarily abundant — but the abundance is bottlenecked through a very small number of carriers. That is not simply distributive inequality. It is dependency architecture.
The NWG end state
The framework’s first move is to see this as a jurisdiction problem rather than a wealth problem, and to separate two things ordinary debate conflates. Technical capacity is what civilization can do in aggregate; ecological capacity is how much differentiated life the civilization can actually support and empower. A society can hold enormous technical capacity and very low distributed capacity — one node doing the work of a hundred million people is not ecological flourishing, it is one node grown unimaginably powerful. So the framework asks the question standard economics often skips: how much of the new capacity is actually available to participants across the field? When new capacity is discovered and then privately gated, the field has found a new source of power and locked it behind a toll — capacity captured rather than distributed. And the capture cascades across the whole living system: gated compute gates signal access; controlling the tools makes faithful embodiment depend on permission; controlling platforms makes standing conditional; controlling infrastructure narrows traversability; requiring proprietary systems makes commensurate participation asymmetric.
So the end state is not “break up monopolies.” It is a structural rule: no private actor should possess effective jurisdiction over functions on which everyone else’s standing depends — foundational compute, identity systems, essential models, payment rails, communication infrastructure, healthcare decision systems, educational access, public knowledge, autonomous logistics. Not because all of these must be state-owned, but because the field cannot permit one actor to become the sole gatekeeper of an indispensable function. Stated positively: indispensable ecological functions require plural access, contestability, or public standing — achievable through competing providers, interoperability, open standards, public infrastructure, cooperative ownership, common compute pools, utility regulation, open-source alternatives, portability, antitrust, or public-interest licenses. The framework does not prescribe the mechanism; it identifies the failure condition, and the failure condition is precise: a participant or institution becomes impossible to route around.
That is why exit carries so much weight here, and why the framework distinguishes formal exit from traversable exit. “Nobody forces you to use our AI” is a fictional freedom when every employer, hospital, school, government, and bank depends on it: exit must not equal ecological death, which means alternatives have to be materially viable, not merely permitted. And because dominance can be earned honestly — one provider genuinely safer, cheaper, better, with society voluntarily routing more and more through it — the framework applies its drift safeguard directly: the stronger a form becomes, the stronger its obligation to remain open to review. A dominant provider should become more contestable as dependency grows, not less, which cuts against normal market logic. From this comes one of the strongest rules in the whole set: ecological importance creates obligations that ordinary ownership does not erase. Once your system becomes infrastructure for the field, you no longer stand where an ordinary private seller stands; the larger the dependency, the larger the stewardship obligation.
The transition gap
The gap is enormous because current property law makes ownership sovereign by default, and concentration is self-accelerating — returns buy more compute, data, talent, and political influence, which buy more returns. Worse, this scenario collides directly with displacement in a dangerous loop: automation weakens labor’s leverage while strengthening owners’, wealth concentrates, alternatives become harder to finance, dependence grows, and owners gain still more jurisdiction. And timing is decisive: plural capacity is far easier to preserve than to recreate, because once one node controls the field, building an alternative may require the very infrastructure that node controls. Living systems do not place every vital function in one fragile node, and the window to keep the field redundant closes as concentration advances.
Transition projects
Like displacement, this is a major transition program, not a single policy, and it favors early anti-concentration architecture — not because wealth is suspect, but because ecological redundancy is protective. The full architecture is on its own page: the transition program, seventeen projects across five programs — measure concentration (an ecological-concentration observatory, dependency-and-exit audits, political-influence audits, vertical-capture mapping, and transfer accounting for ownership); preserve alternative capacity (public/civic compute, open models and standards, a model commons, small-institution access, ecological redundancy); preserve exit and forkability (data portability, forkability, procurement diversification); redistribute capacity (an automation dividend and capacity return, distributed-ownership experiments); and limit jurisdiction (ecological antitrust, critical-infrastructure classification) — sequenced from now to a mature state.
Capture risks
The signature capture is nominal redistribution that leaves real control untouched — “stakeholder” language, token equity, advisory boards — while decision rights stay concentrated. A second is competence-laundering: “they won because they’re best” used to convert genuine excellence into permanent, unquestionable dependency. A third is vertical capture, where power in one layer (chips → cloud → model → agent → marketplace → payment → identity) is used to acquire jurisdiction in the next. A fourth is the state itself amplifying concentration through single-provider procurement. And a fifth is speed laundering — promising the end state at a horizon comfortably far enough away that concentration completes before anything binds.
The limits
This is where the framework is clearest about what should change and weakest on how to compel it. NWG can say standing must not depend on market usefulness; it cannot, by itself, declare that wealth will therefore be shared — there has to be an actual political economy underneath (taxation, ownership reform, public dividends, public equity, sovereign wealth, cooperatives, or some mix), and the beneficiaries of concentration may rationally oppose anything that reduces their jurisdiction. So the real question is not what would a healthy system look like? but why would powerful owners agree to it? — and the framework does not have a sufficient answer. Its position: a strong normative diagnosis, a weaker transition mechanism, and a hard boundary once concentration has already captured lobbying, capital, compute, infrastructure, and security — at which point NWG can diagnose the failure beautifully and cannot necessarily reverse it. That boundary is exactly where this scenario passes into gatekeeper AI, where concentration moves from ownership into the everyday decision architecture of people’s lives.
The hard standard
Standing is defended in principle — the framework insists it be independent of ownership — but this scenario exposes the material catch: declaring universal standing does not answer who controls the resources required to make that standing real. Ecological capacity is the leg most directly attacked, and in the exact form the framework warns about: technical capacity can soar while distributed capacity falls, capacity discovered and then gated — the captured-capacity failure, output up and ecological capacity hollow. Ontological correctability turns on whether the field can still route around a dominant node: forkability, plural providers, and real exit are what keep a dominant form contestable, and their loss is what lets a merely-successful arrangement harden into one that cannot be reopened.
The framework belongs in the strong class because its diagnosis reframes the whole problem correctly — from how much inequality is fair? to how much indispensable jurisdiction has accumulated in one node? — and because private-versus-public turns out to be the wrong axis: a private firm can be tolerable if it is interoperable, exitable, and routable, while a widely-held public company can be dangerous if it is impossible to bypass. The node genuinely at risk is capacity-as-distribution, and the limit is not conceptual but political: the framework can specify plural access, real exit, and stewardship obligations, and cannot by itself supply the power to make concentrated owners surrender jurisdiction once they hold it.
The goal is not to prevent powerful institutions from emerging. It is to prevent powerful institutions from becoming impossible to route around.
Where this tends to land hardest — in the regional vulnerability map: the United States.