The ownership-concentration scenario is best treated as a major transition program, because concentration can quietly turn every other scenario into a jurisdiction problem. The threat is not that some people become extremely wealthy; it is that a few actors gain control over the infrastructures through which everyone else must think, build, transact, communicate, and participate. So the projects below address ownership, access, dependency, contestability, jurisdiction, and alternative capacity together — and they favor early action, because plural capacity is far easier to preserve than to recreate once one node controls the field.

A note on reading this: the framework is not anti-wealth and not automatically anti-scale. A company becoming huge may reflect genuine success. The problem begins only when scale becomes unavoidable jurisdiction — when a system becomes impossible to route around.

A. Measure concentration

Ecological concentration observatory. Measure not just market share but ecological dependence — concentration across frontier models, compute, chips, cloud, data centers, training data, capital, distribution platforms, identity systems, payment rails, enterprise software, robotics, agents, energy contracts, and developer ecosystems, together with switching costs, interoperability, the ability to self-host, and the concentration of critical talent and standards. A market can look competitive while every firm in it depends on the same few underlying providers, so the real question is how much of the ecology can no longer route around this node?

Dependency and exit audit. For each critical dependency, determine whether the exit is real or merely contractual: how long switching would take, what it would cost, whether data would transfer, whether operations could continue, whether a technical alternative actually exists, and whether in-house knowledge remains — classifying each as easy, difficult, nominal, or effectively no exit. Competition is not meaningful if switching is practically impossible.

Political-influence concentration audit. Track whether AI wealth converts into disproportionate political jurisdiction — lobbying, donations, regulatory staffing pipelines, procurement dependence, advisory and standards-body influence, media reach. Economic concentration becomes self-reinforcing once owners can shape the laws governing their own infrastructure.

Vertical-capture mapping. Map when one actor controls too many layers at once — chips → cloud → model → agent → marketplace → payment → identity — producing bottleneck analysis, self-preferencing rules, and, where necessary, separation requirements. A firm may use power in one layer to acquire jurisdiction in the next; the structural-separation remedies this mapping justifies belong to the limit jurisdiction front below.

Ecological transfer accounting for ownership. Apply ecological transfer accounting to concentrated firms: what capacity they absorb (capital, data, compute, talent, energy), what they create (productivity, research, tools), what they distribute (open access, lowered costs, public benefit, new businesses), what they withhold (proprietary bottlenecks, switching costs, exclusive licensing), and what burden they export (labor displacement, systemic dependency, political concentration, cyber and energy strain). The central question: is this node increasing global ecological capacity, or merely accumulating ecological capacity inside itself?

B. Preserve alternative capacity

Public / civic compute. Preserve at least one viable compute pathway outside concentrated private ownership — public clusters, university and cooperative pools, municipal or regional infrastructure, nonprofit capacity, public-interest cloud credits. This need not replace private infrastructure; it provides a route that still exists when private access is denied, overpriced, or strategically restricted. Without it, “open competition” can be fictional.

Open models and open standards. Prevent the ecology from becoming dependent on proprietary formats and closed interfaces — open model ecosystems, interoperable APIs, portable workflows, standard agent protocols, model portability, open safety and evaluation tooling. The goal is not that every frontier model be open-source; it is that participants not become permanently trapped inside one provider’s architecture, since the ability to fork is central to correctability.

Model commons / public knowledge. Keep some foundational capabilities available as a commons — public datasets, evaluation tools, safety research, educational and scientific models, public-domain training resources. If all fundamental AI capability sits behind private gates, civic institutions become permanent renters of cognitive infrastructure.

Small-institution access. Make advanced AI usable by more than large organizations — compute credits, technical assistance, shared agents, procurement support, public APIs — for small businesses, schools, nonprofits, municipalities, community organizations, and independent researchers. Concentration is not only about who owns capability but about who can operationalize it.

Ecological redundancy. Maintain multiple viable carriers for critical functions — independent providers, technical, geographic, ownership, and architecture diversity, plus human fallback capacity. Concentrated systems are efficient but brittle, and a healthy ecology may deliberately tolerate some inefficiency in exchange for redundancy, exactly as living systems avoid placing every vital function in one fragile node.

C. Preserve exit and forkability

Data portability and jurisdiction. Prevent providers from turning accumulated user data into irreversible dependence — participants able to export data, move history, context, agent configurations, and business workflows, revoke access, and understand what is retained — with limits on using one domain’s data to acquire jurisdiction in another. A provider can become dominant not because its model is best but because leaving means abandoning years of accumulated context.

Forkability. Ensure important systems can generate legitimate alternatives — communities able to fork software, organizations to adopt competing standards, developers to build compatible alternatives, users to carry data and identity elsewhere. Forkability is a resilience property, because systems that cannot fork tend to become ontologically rigid.

Public procurement diversification. Keep the state from accidentally amplifying concentration: if every agency buys from the same provider, procurement becomes a monopoly amplifier. Require multi-provider strategies, open standards, portability, backup providers, contract exit clauses, and public fallback capability, so the state does not build the concentration it later tries to regulate.

D. Redistribute capacity

AI wealth / capacity return. Determine how the extraordinary returns of concentrated AI ownership return to the wider ecology — public equity, sovereign wealth funds, automation dividends, cooperative and employee ownership, public-interest licensing, tax structures, community-investment obligations. No single mechanism need be chosen first; the principle is that if the ecology supplies the data, infrastructure, labor, education, law, energy, and stability that make AI wealth possible, some portion of the resulting capacity should circulate back into the ecology. This front is shared directly with the displacement program.

Distributed-ownership experiments. Prototype alternatives before a crisis forces them — user-owned AI cooperatives, worker-owned model services, community data trusts, public-private utilities, cooperative agent platforms, local compute cooperatives — as experiments rather than ideological commitments, so we learn empirically which ownership structures distribute capacity without destroying performance.

E. Limit jurisdiction

Ecological antitrust. Reframe antitrust around jurisdiction rather than price: instead of are prices too high?, ask whether one node has accumulated too much indispensable jurisdiction — inability to route around it, control of essential interfaces, suppressed emergence of alternatives, high switching costs, control of standards, downstream self-preferencing, ability to exclude competitors from necessary infrastructure. Many digital monopolies offer low prices while becoming ecologically dominant, which price-based antitrust cannot see.

Critical-infrastructure classification. Identify when a system stops being an ordinary private product and becomes critical ecological infrastructure — a large share of society depends on it, its failure would disrupt essential services, no practical substitute exists, many downstream systems require it, and refusing access materially harms participation. Past that line, ordinary ownership no longer implies unlimited jurisdiction, because ecological importance creates obligations that ordinary ownership does not erase.

The sequence

The projects are staged by how far concentration has advanced, not by fixed dates — and the early ones matter most, because redundancy is far cheaper to keep than to rebuild.

Phase What gets built
Now Concentration observatory; dependency mapping; open standards; portability requirements; public-compute pilots; procurement diversification.
Early concentration Infrastructure classification; ecological antitrust; distributed-ownership pilots; capacity-return mechanisms.
High concentration Structural separation where necessary; utility-style obligations for critical infrastructure; stronger public alternatives; hard interoperability mandates.
Mature NWG state Private ownership can remain substantial — but no participant is allowed to become the only viable route through an indispensable ecological function.

That final line is the destination principle. Ownership concentration becomes a problem worth all this machinery only when concentrated ownership begins to determine who can access essential capacity, which alternatives can exist, and whether participants can meaningfully exit.

The goal is not to prevent powerful institutions from emerging. It is to prevent powerful institutions from becoming impossible to route around.