This publication has completed its full initial draft — all twenty sections, the ICRI/SICRI/EEC construct panel, the rights framework, scenario comparison, and closing sequence are below. It has not yet been through the Foundation's full hostile-review process (the multi-round adversarial review RP04 and the CIRE Working Paper both underwent before final publication), nor has it received the complete claim ledger, source audit, or figure-production pass the original specification called for. Treat this as a complete, internally-consistent working draft ready for external review, not yet a fully finalized Foundation publication.
The Foundation's prior publication, When Work Is No Longer the Ticket to the Economy, asked what replaces employment once wages no longer scale with productive output, and proposed that distribution — not production — is the constraint AI is placing under strain.
That paper left a question unanswered by design: even if a replacement distribution mechanism is built — an automation dividend, a sovereign fund, universal capital ownership — who controls it, on what terms, and what happens to a person's participation in the economy if that control is exercised against them? A currency or credit system can solve the demand-side arithmetic in RP04's identity while still being an instrument of domination. This paper is about that second, harder question: not how purchasing power is distributed, but who governs the systems that distribute it, and what protects a person's ability to actually use what they're given.
Artificial intelligence and robotics may progressively reduce the human labor required to produce goods, services, and infrastructure. The resulting problem is usually framed as unemployment. That framing is incomplete. If productive capacity becomes increasingly automated, the more consequential question is who owns that capacity, who receives claims on its output, and which institutions decide the conditions under which people may exercise those claims.
This paper examines four ownership architectures for a highly automated economy — concentrated private ownership, centralized public ownership, universal distributed citizen ownership, and community-scale cooperative ownership — and argues that none is sufficient alone. It evaluates programmable currency, consumption-credit systems, and demurrage, concluding that replacing dollars with tokens does not by itself solve the post-labor distribution problem: currency is a representation of a claim, and the decisive questions are how claims are created, who receives them, whether they can be revoked, and what constrains the issuer. It introduces a formal Claim Governance Lifecycle and a three-construct measurement panel — the Institutional Convergence Risk Index (ICRI) and its systemic extension (SICRI), measuring how much authority over production, identity, currency, surveillance, and enforcement has concentrated in one institution or class of owners; and Economic Exit Capacity (EEC), measuring the alternatives available to a person if that authority is exercised against them. Using OECD institutional-ownership data and 2025–2026 SEC filings, it shows that claims about a unified billionaire class owning the automation stack outright are not supported by the evidence, while a narrower, still-serious claim is: voting control, operational control, and institutional asset concentration are already high in the firms building frontier automation, independent of majority economic ownership. Its governing principle: no institution controlling essential automated productive capacity should hold unilateral authority over production, identity, surveillance, currency, communication, and enforcement at once. The paper does not claim a coercive outcome is intended, planned, or inevitable — only that existing concentration is not starting from a neutral baseline, and that diffusing the governance authority automation generates is a design choice, not an automatic byproduct of growth.
The first post-labor question, addressed in RP04, is how purchasing power reaches people once wages stop scaling with output. The second question — this paper's subject — is who controls the system that performs the replacement.
Consider a universal monthly payment distributed through a programmable digital currency. At first glance this appears to solve the participation problem: citizens receive purchasing power without employment and can continue consuming the output of an automated economy. The description is incomplete. Who issues the payment? Who determines its amount? Can it be spent everywhere, or only on approved categories? Can it be saved, transferred, or inherited? Does it expire? Can it be frozen following a criminal conviction, a civil accusation, an administrative error, a political protest, a tax dispute, or a platform terms-of-service violation? Can a person leave one jurisdiction or provider and retain access? Can the issuer observe every purchase, and can it prevent purchases of disfavored goods?
These are not secondary implementation details. They determine whether a distribution mechanism delivers freedom or merely delivers conditional permission. The post-labor problem therefore has two dimensions, not one:
A viable post-labor system needs both the claim and its usability satisfied. A society can preserve consumer demand through universal digital payments while simultaneously building a highly coercive economic structure; it can nationalize AI in the name of equality while concentrating unprecedented informational and economic power in the state; it can leave automation in private hands while taxing a portion of the gains, even as the owners of that automation accumulate enough political influence to shape the tax rules that constrain them. The economic distribution problem and the institutional power problem are related but not identical, and this paper treats them as such throughout.
Discussions of economic claims tend to collapse three distinct forms of control into a single word — "control" — that in practice separate cleanly and create different risks depending on which institution holds which:
These three powers are frequently assumed to travel together, but they need not. A government might allocate claims broadly and progressively while a separate, privately-operated payment platform holds usage control through terms-of-service restrictions the government never authorized. A corporation might hold no allocation power at all — it issues no benefits — yet hold decisive usage control simply by operating the infrastructure a claim must pass through to be spent. A court might hold formal revocation authority while an administrative agency or automated system exercises it in practice, with the gap between formal and actual authority becoming the point at which due process fails. This three-way distinction recurs through the rest of this paper, including in the design of the ICRI construct in Section VII, and the Foundation regards it as a necessary corrective to treating "who controls the money" as a single, undifferentiated question.
This paper's analysis is built primarily on objectively verifiable conditions — enacted statutes, documented institutional structures, and formally specified constructs — and on falsifiable hypotheses derived from them. A third category recurs throughout, and the Foundation states its methodological status explicitly rather than leaving it implicit: what a population believes about the fairness, trustworthiness, or coercive potential of an economic system is analytically distinct from whether that belief is factually accurate, but it is not thereby irrelevant. Widely-held distrust of concentrated economic power, or anxiety about automation-driven exclusion, shapes voting behavior, market confidence, regulatory pressure, migration decisions, and civil compliance regardless of how precisely that distrust tracks the underlying institutional reality. The Foundation therefore treats public perception neither as evidence that a given proposition about institutional risk is true, nor as noise to be filtered out before real analysis begins, but as a distinct, empirically observable variable in its own right — measurable through existing survey and public-opinion research on institutional trust — that influences the durability and adoption of any governance mechanism this paper or RP04 proposes, independent of that mechanism's technical merit.
A worked example of the translation this note describes: the intuition that concentrated wealth-holders are "betting on a world that won't exist" does not survive as a factual claim — Section II's identity shows capital claims (Ki) remain fully meaningful under automation, and may concentrate further rather than evaporate. Its research-legitimate form is narrower: current investment incentives remain largely organized around maximizing returns within labor-mediated market structures, even as AI may progressively weaken the coupling between labor and production; whether existing ownership incentives remain stable under widespread automation is accordingly an open institutional question this paper treats empirically (Sections V–VII) rather than assumes in either direction.
RP04 developed purchasing power as its central measured quantity. This paper has so far developed claim usability (Section I above) and institutional convergence (Section VII). A society's economic institutions do not persist merely because they are well designed by the criteria this paper applies — they persist because enough of the population affected by them regards them as legitimate enough to participate in voluntarily rather than merely comply with under duress. Institutional legitimacy therefore depends not only on whether governance mechanisms are objectively well-constrained, but on whether they are perceived as fair by the people expected to rely on them, and that perception is itself an empirical variable influencing adoption, compliance, political stability, and institutional durability — regardless of whether every underlying perception is factually correct. The Foundation flags institutional legitimacy as a candidate fourth pillar alongside purchasing power, claim usability, and convergence risk, and — consistent with its practice of not rushing unvalidated instruments into use — does not attempt to formalize it as a fifth measurement construct in this draft. Whether it warrants its own construct, or is better treated as a variable feeding into EEC and a future public-trust indicator, is left as an open question for the research agenda.
A credit-token system could technically replace some functions dollars currently perform. The token itself is not the solution to the problem this paper is organized around, and the reasoning is worth making precise rather than asserted. It rests on separating three layers that public discussion of "replacing the dollar" routinely collapses into one:
PRODUCTION → ECONOMIC CLAIMS → CURRENCY
Production creates value. Claims determine who is socially and legally entitled to a portion of that value. Currency is merely the instrument that records and transfers those claims. A proposal to replace dollars with tokens is, by this layering, a proposal about the third layer only — it leaves the first two layers, which is where this paper's actual argument lives, untouched by construction.
An economic claim is a socially recognized entitlement to command some portion of present or future production, whether arising through wages, ownership, statutory right, transfer, debt, or another legally recognized mechanism. Currency is one instrument through which such claims are represented and exchanged. It is not itself the source of the claim's legitimacy, and a change in currency technology does not, by itself, change who holds a claim, how much, or under what conditions.
Formally, for a given person i:
Most existing treatments of post-labor income group unemployment insurance, means-tested welfare, pensions, and universal basic income together under a single "transfers" heading. This paper separates them by the source of legitimacy each claims. A conditional transfer (Ti) is granted on the basis of a tested condition — unemployment status, income level, disability determination — and can, by the same logic that granted it, be withdrawn when the condition changes or is judged no longer met. A rights-based distribution (Ri) derives its legitimacy from citizenship or recognized legal personhood rather than demonstrated need or economic contribution — a universal basic income or a citizen equity dividend granted to everyone as a matter of status, not circumstance. The distinction matters because the two categories differ sharply in how defensibly they can be revoked, which is the central governance question the rest of this paper is built around, and which prior participation-economics literature — including RP04's own eleven mechanisms, evaluated primarily on funding and political tractability — did not examine on these terms.
Every economic claim, regardless of which of the five terms above created it, passes through a common sequence of governance points. The Foundation proposes this sequence as an organizing framework used throughout the rest of this paper, not as a new mathematical construct in its own right.
CREATION → ALLOCATION → RECOGNITION → EXERCISE → CONVERSION → TRANSFER → MODIFICATION → REVOCATION → APPEAL → EXIT
The question this reframes is not "what currency replaces the dollar?" It is "what rule creates a legitimate and durable claim on automated production for a person no longer required as labor, and what governs that claim at each stage of its lifecycle thereafter?" The remainder of this paper is organized around that second half.Level B — Follows directly from RP04's own framework once the claim-formation question is separated from the claim-governance question
A consumption-credit system — periodic issuance of digital credits, accepted for goods and services, funded through taxation, public ownership income, sovereign fund returns, or automation levies — is technically feasible. The technology is not the constraint, and the evidence for that single proposition is what the following two data points are offered to establish — this section is not a survey of CBDC policy for its own sake. Central bank digital currency infrastructure now exists in mature, operational form in multiple jurisdictions: as of an April 2026 industry tracking survey, five CBDCs were already in live retail or quasi-retail production (Nigeria's eNaira, the Bahamas' Sand Dollar, Jamaica's JAM-DEX, the Eastern Caribbean's DCash, and China's e-CNY), with roughly forty more in pilot — figures that vary by tracker depending on how "live" and "pilot" are defined, and are cited here as an order-of-magnitude indicator rather than a precise count.1 The United States took a different path: Pub. L. 119-101, Section 1001 (the CBDC provision of H.R. 6644, the 21st Century ROAD to Housing Act, a housing-affordability bill), enacted July 2026 after passing the Senate 85–5 and the House 358–32, imposes a four-year statutory prohibition on Federal Reserve CBDC issuance through December 31, 2030, driven by lawmakers' financial-privacy and surveillance objections rather than any technical infeasibility finding — the enacted statutory text is the primary source for this claim, with the NLIHC and Bipartisan Policy Center section-by-section explainers cited as secondary confirmation of the section's content and purpose.2 This is a separate and later measure than the House-passed but Senate-stalled Anti-CBDC Surveillance State Act (H.R. 1919, July 2025) — a distinction worth stating explicitly, since the two bills are easily conflated and only the later provision, inside the housing bill, has actually been enacted.3 The global range, from live retail CBDCs to a Congress that just voted by wide bipartisan margins to statutorily forbid one, is itself evidence that the open question is institutional trust, not engineering — which is the single point this section exists to establish before returning to governance for the remainder of the paper.
The potential advantages of a consumption-credit system are real: rapid distribution during economic shocks (the 2020–2021 US stimulus payments moved on the order of $800 billion through existing rails within weeks, demonstrating that plumbing is not the bottleneck once political will exists4), reduced administrative complexity relative to fragmented means-tested programs, and direct support for aggregate demand during an automation-driven transition of the kind RP04 models. The risks are equally concrete, and the Foundation states them with the same weight: universal transaction surveillance, political conditioning of access, purchase restrictions on lawful goods, revocation without meaningful due process, dependency on a single issuer or technical network, and — the risk this paper treats as most structurally important — the conversion of economic citizenship into behavioral compliance, where continued access to one's own claim depends on ongoing approval of one's conduct.
Every capability that makes a programmable credit useful for legitimate purposes — fraud prevention, automatic emergency relief, disaster-targeted assistance, child-support enforcement — is the same capability that is capable of coercive application. This is not a flaw specific to any one government or company's implementation; it is a property of programmability itself, and stating it this way is deliberate: the claim is about technical capacity, not about the intent of any current issuer. A system cannot be granted the power to selectively restrict spending for good reasons without also being granted the technical capacity to restrict it for bad ones. The design question is therefore not whether to build programmability, but what institutional and legal separations constrain who can invoke it, under what published rules, and with what appeal — the Appeal and Revocation stages of the Claim Governance Lifecycle in Figure 1.
The Foundation's working principle, offered for scrutiny rather than as a settled conclusion: no essential medium of economic participation should be revocable, conditionable, or comprehensively surveilled without transparent law, meaningful due process, institutional separation between the issuer and any enforcement body, and an accessible non-programmable fallback. This is a design criterion for evaluating any of the eleven mechanisms RP04 surveys, not a claim that programmable currency is illegitimate as such — cash itself was, in its origin, a technology that reduced a different kind of dependency (barter and personal credit relationships), and the Foundation does not treat "new payment technology" as inherently more dangerous than what it replaces, only as requiring the same scrutiny any concentration of control over a population's economic access would warrant regardless of its technical implementation.
Demurrage — a mechanism by which currency or credits lose value over time, discouraging hoarding — has a genuine intellectual lineage. Silvio Gesell's early-twentieth-century Freiwirtschaft ("free economy") theory argued that money's ability to be hoarded without cost gave it an unfair advantage over goods, which naturally depreciate, and proposed a stamped scrip that lost value unless a stamp was periodically purchased to maintain it.5 The theory was tested, briefly and at genuinely small scale, in Wörgl, Austria, in 1932–1933: the town issued labor certificates that lost 1% of their value monthly unless stamped, reportedly increasing local spending velocity and public-works employment during the Depression, before the Austrian National Bank asserted its currency-issuance monopoly and shut the program down after roughly thirteen months.6 The Foundation states plainly what this case does and does not establish: it is real, documented, and shows demurrage can function at a single-town scale for about a year under Depression-era conditions with a fixed, small, cooperative population. It does not establish that the mechanism scales to a modern national economy, survives normal (non-crisis) conditions, or avoids the distributional problems described below.
Forcing velocity is not automatically welfare-improving, and this is the paper's central objection to treating demurrage as a general-purpose solution rather than a targeted instrument. People save for reasons a pure consumption-credit design does not accommodate: emergencies, housing down payments, education, relocation, and retirement. A system in which all distributed claims expire could trap recipients in continuous short-term consumption while the ownership class — holding capital assets rather than expiring credits — continues accumulating durable wealth unaffected by demurrage. This produces a two-class risk the Foundation regards as a serious, underexamined objection to demurrage-based UBI proposals: the majority receives expiring consumption claims while an ownership minority retains permanent, appreciating capital claims. In RP04's own terms, that pattern could satisfy the nominal purchasing-power condition PtN ≥ YtN (aggregate demand, in nominal terms, meeting nominal output) while doing nothing to address the capital-concentration dynamic RP04 Section 6.2 separately identifies — a reminder that RP04's aggregate condition and this paper's distributional concern are answering different questions, and satisfying one does not satisfy the other.Level C
A more defensible design, consistent with the Foundation's practice of proposing structure rather than a single instrument, would separate at minimum: baseline consumption credits (potentially subject to limited, bounded demurrage); protected personal savings, exempt from decay up to a defined threshold; a distinct mechanism for acquiring durable capital participation, of the kind examined in Section 5.3 below; and emergency reserves inaccessible to ordinary decay rules. Demurrage, on this view, is a targeted velocity instrument evaluable alongside RP04's eleven mechanisms — the Foundation has not identified a substantial contemporary literature establishing demurrage as a sufficient, economy-wide replacement for ordinary money, savings instruments, and capital ownership, and evaluates it here as a targeted monetary-design option rather than a complete post-labor architecture.
Four architectures for owning automated productive capacity are evaluated here on the same basis: none is assessed as inherently good or bad, but on where its risks concentrate.
Corporations and investors retain control over automated infrastructure; governments preserve broad participation through taxation and transfer, per RP04's Section IX mechanisms. Advantages: strong investment incentive, decentralized entrepreneurial decision-making, rapid innovation under contestable competition, and continuity with existing legal and corporate-governance institutions — the same institutions RP04 Section III examined as rational actors, not villains. Risks: capital-income concentration (RP04 Section 6.2), regulatory capture (examined in a future section of this paper), and the specific risk this paper adds to RP04's treatment — that essential infrastructure control by a small number of firms creates political influence derived from economic dependency, independent of any tax policy applied to their profits. The model's compatibility with broad economic agency is likely to depend heavily on whether competition law, data rights, and credible public or interoperable alternatives remain strong enough to constrain concentration; the Foundation does not assume that condition holds by default, and treats it as an empirical question addressed partially in a planned regulatory-capture section. A nominal public alternative that is technically inferior or practically inaccessible would not provide meaningful constraint, which is why "credible" is doing real work in that sentence.
The state owns or controls foundational AI systems, energy networks, and major automated production. Advantages: public capture of automation gains, universal-service obligations, coordinated infrastructure investment, reduced private monopoly rents. Risks: politicized resource allocation, integration of distribution infrastructure with state surveillance and identity systems, bureaucratic stagnation, and — the risk most directly relevant to this paper's governance principle — concentration of productive and coercive power within the same institutional order, where the constraining role ordinarily attributed to market entry and competition must instead be performed by constitutional separation, judicial review, political competition, independent oversight, and meaningful rights of exit. Those checks are real institutions with a genuine track record, not merely theoretical — but they can also fail, weaken, or be captured, and the Foundation treats their adequacy in any given jurisdiction as an empirical question rather than assuming public ownership is either safer or more dangerous than private ownership as a category. Public ownership can reduce private domination while increasing state domination, or it can do neither if its governance is genuinely plural, transparent, contestable, and legally constrained; the relevant question is the governance design, not the ownership label.
Every citizen receives a durable interest in automated productive capacity — shares in a national automation fund, non-transferable beneficial ownership units, or a public trust holding automation-derived royalties. This is RP04's "citizen equity" mechanism (Section 9.2/9.9) examined here specifically for its governance properties rather than its funding tractability. Advantages: direct alignment between automation gains and household income, reduced wage dependency, and the possibility of broadening capital ownership through asset allocation rather than relying exclusively on recurring cash transfers — though capitalizing the underlying fund still requires an identified source (transferred public assets, purchased equity, newly issued ownership interests, automation-license revenue, or taxation; the mechanism does not avoid this question, only potentially changes its form). Risks: distribution volatility tied to market performance, fund-governance capture, and — the risk this paper adds that RP04's funding-focused treatment did not examine — reconcentration through sale, inheritance, or predatory lending against the equity claim, though this risk is conditional on the specific ownership form chosen: if units are universally reissued each period or held beneficially in a perpetual public trust rather than personally owned and inheritable, several of these reconcentration pathways do not apply. A durable design would likely require some portion of the claim to be non-transferable and non-collateralizable. This yields a testable hypothesis rather than a mere caution: systems permitting unrestricted sale or collateralization of universal citizen equity will exhibit materially faster ownership reconcentration than systems preserving a protected, non-transferable core — a proposition the Foundation flags for future empirical work rather than treats as established here.
Automated productive systems held under municipal, cooperative, tribal, or neighborhood governance — community energy grids, cooperative agricultural robotics, local fabrication systems, regional data trusts. Advantages: resilience, local accountability, additional institutional alternatives that can shorten lines of accountability, and compatibility with regional preference diversity. Risks: limited access to frontier-scale capital, uneven capacity across communities of different wealth, local corruption or exclusion at smaller scale, and technological fragmentation. Exit capacity specifically is not uniformly better at the local scale than the national one — leaving a captured local cooperative can require selling a home, changing schools, moving away from family or community ties, or relocating employment, costs a national system with genuinely interoperable alternatives may not impose. Community ownership is unlikely to substitute fully for national or global-scale systems in its current form — the Foundation has not identified a community- or worker-owned cooperative independently financing and operating a frontier-scale model-training program comparable to those of the largest commercial and state-backed laboratories as of this writing — but its distinct value, developed further in Section VI, may be as a structural counterweight: ensuring some minimum productive capacity remains outside both centralized private and centralized public control.
| Architecture | Distribution Breadth | Innovation Potential | Surveillance Risk | Exit Capacity | Capture Risk |
|---|---|---|---|---|---|
| Concentrated private | Low unless taxed | High | Medium–High | Medium | High |
| Centralized public | Potentially high | Medium | High | Low | High |
| Universal distributed | High | Medium–High | Medium | Medium–High | Medium |
| Community/cooperative | Medium–High, locally | Medium | Low–Medium | High | Medium, locally |
Ordinal, qualitative scoring, not a verdict. These ratings describe tendencies under relatively concentrated, unconstrained versions of each architecture — not inherent, fixed properties of the ownership label itself. Each score is conditional on market structure, constitutional safeguards, interoperability, financing, and governance design; a well-constrained version of any architecture in this table could score materially better than shown, and a captured or poorly-designed version of any architecture could score worse. The Foundation has not attempted to manufacture false quantitative precision where none currently exists in the underlying research.
No pure architecture in Section V scores well across all five dimensions. The Foundation proposes, as a candidate for comparative evaluation rather than a recommended final system, a mixed plural architecture along the following lines: privately developed competitive applications; foundational infrastructure governed through some combination of public regulation, open standards, public-interest obligations, and, where justified, limited public or shared ownership — deliberately not specifying "publicly regulated but not publicly owned," since Section 5.2 established characteristic risks of full centralized public ownership of the entire productive system, not that any public stake in foundational infrastructure is categorically inferior; a broadly distributed and legally protected beneficial interest in a defined portion of automation-derived capital, with transfer, collateralization, inheritance, and creditor-access rules designed to prevent rapid reconcentration — leaving open, rather than presupposing, questions this paper has not resolved: individual accounts versus collective trusts, citizens versus lawful residents, and the treatment of inheritance, divorce, and migration; community ownership of essential local systems; open technical standards enabling switching between providers; and — the requirement this paper treats as load-bearing — strict institutional separation between currency administration and any surveillance or enforcement function.
The Foundation is not hypothesizing that this architecture is universally "best" — only that it may produce a more favorable multi-objective risk profile than architectures concentrating authority in a single ownership form. Stated as a testable proposition:
H1: Under comparable technological and fiscal conditions, an institutional architecture distributing control across private, public, citizen, and community entities will exhibit lower convergence risk (Section VII) and greater exit capacity than architectures concentrating productive and claim-governance authority predominantly within one ownership form.
H2 (counter-hypothesis): The plural architecture's coordination costs, accountability gaps, and interoperability failures will outweigh its concentration-risk benefits in practice.
H3 (counter-hypothesis): Ownership plurality alone will not reduce convergence risk if the nominally separate participating entities share identity systems, payment rails, data infrastructure, or common administrative control — multiple nominal owners can still function as a single integrated system, which is precisely the unit-of-analysis problem Section VII's ICRI construct is built to address.
Section 5's table already shows every pure architecture concentrating at least one serious risk that a different architecture happens to mitigate; that pattern is what motivates H1, and H2/H3 are stated here specifically so this paper cannot later claim the mixed architecture succeeded by only ever testing it against H1. A future scenario comparison (planned) is intended to evaluate H1 against H2 and H3 directly rather than leave the candidate architecture's superiority as a qualitative judgment.
The Foundation introduces a new construct, in the tradition of RP04's LRT/EPI/PE/DRR panel. Like those four, it is offered as a research program requiring substantial validation work, not a finished instrument. This section went through an internal revision after an initial version was found to double-count interactions and conflate control, integration, and coercive discretion into a single insufficiently specified term — the corrected version below reflects that revision, stated rather than hidden, consistent with the Foundation's practice of preserving negative results.
The highest-risk configuration this paper identifies is not any single technology or ownership form in isolation. It is one institution — or a functionally coordinated network of institutions, defined precisely in 7.4 below — controlling several essential domains simultaneously. The risk is not additive across domains; it is interactive. Controlling identity alone is a real but bounded risk. Controlling identity and currency together is materially more coercively powerful than either independently, because it becomes possible to condition a person's ability to transact on who they are proven to be. Controlling currency, surveillance, and enforcement together is more powerful still, because the system that observes behavior, the system that controls economic access, and the system with physical enforcement capacity become the same actor.
"Production" as a single domain is too broad to score meaningfully — it could mean one factory, a national food system, or a dominant marketplace, each with a different risk profile. The Foundation proposes a more bounded initial taxonomy, offered as a starting point for revision rather than a closed list: foundational compute; model or decision infrastructure; essential-goods production; energy; digital identity; payment and settlement; behavioral data; communications; essential service access (housing, healthcare, transportation); civilian enforcement; and military or autonomous force. Each domain is scored not by an institution's mere presence in it, but by its dependency-weighted control:
The paper has so far referred loosely to "one institution or a tightly coordinated network," which invites manipulation in either direction — a researcher could score nominally separate companies independently even when they operate an integrated ecosystem, or aggregate unrelated organizations into an invented network to inflate a score. The Foundation proposes a preliminary rule: the assessed entity may be a single legal institution or a functionally coordinated institutional network. A network is treated as a combined unit only when common ownership or ultimate control, shared board membership, contractual exclusivity, a shared technical backend, legally mandated data exchange, a common identity provider, interoperable enforcement access, coordinated decision rules, state direction, or exclusive dependency on one cloud or payment infrastructure permits a decision in one domain to materially affect access in another. This rule is itself a research object rather than a settled test — the Foundation expects it to require refinement once applied to real cases.
The corrected formula's dij and sij terms are the primary defense against false positives, but a further discipline is needed: the index should assess capability, scope, discretion, safeguards, reversibility, and observed use — not an institution's stated purpose. An institution may hold a benign stated purpose while retaining dangerous capabilities; conversely, a lawful anti-fraud purpose may generate temporary, reversible harm without representing systemic domination. Relevant distinctions for scoring dij and sij in practice include the scope of data actually exchanged, whether decisions are automated, whether the institution may act without external authorization, the duration and breadth of any exclusion, the availability and speed of appeal, the reversibility of the action taken, auditability, and the existence of practical alternatives. Stated purpose alone is not sufficient to determine risk, and the construct is designed specifically not to rely on it as the primary signal.
ICRI is planned to sit alongside a second construct, Economic Exit Capacity (EEC — addressed in a future section), and the two are deliberately not duplicates. ICRI measures concentrated capability and discretion held by institutions and systems. EEC is intended to measure the alternatives and protections available to the affected person or community once that capability is exercised against them. An institution can score high on ICRI while individuals retain moderate EEC because robust alternatives exist elsewhere in the system; conversely, a modestly integrated institution can create severe harm where affected people have no practical alternative at all. A future combined exposure measure — provisionally, Convergence Exposure Risk, CER = ICRI × (1 − EEC), assuming both are normalized to a common scale — is flagged here as a direction worth preserving once EEC itself has been developed, not introduced as a formal construct in this draft.
The ICRI's governance implication follows from the interaction term rather than from any specific numeric threshold, which has not yet been established: no institution controlling essential automated productive capacity should possess unilateral authority simultaneously over production, identity, surveillance, currency, communication, and enforcement. This is offered as a design constraint comparable in kind to separation of powers; historical and contemporary financial-structure rules that separate selected banking, securities, custody, clearing, and risk-bearing functions (of which the Glass-Steagall-era commercial/investment banking separation is a partial and, since substantially repealed, imperfect example — a caveat the Foundation states because the analogy is real but should not be read as an uncomplicated success story); antitrust and sector-specific restrictions directed at vertical integration, tying, discriminatory access, cross-market leverage, and excessive concentration; and civilian control of military authority. Each is an existing institutional precedent for the general principle that concentrating otherwise-legitimate powers in one actor creates a risk the powers do not individually present. The Foundation is not proposing this principle as novel; it is proposing ICRI as a way to measure how close a given real institution — public or private — sits to that convergence, so the principle can be applied to specific, evaluable cases rather than invoked only in the abstract.
Section 7.1–7.7 defines ICRI at the level of a single institution or a functionally coordinated network (per the 7.4 unit-of-analysis rule). A distinct and, the Foundation now judges, equally important question is whether convergence risk can accumulate at the level of a class of owners across nominally separate, non-coordinated institutions — a possibility the single-institution ICRI is not designed to detect and could systematically miss.
The relevant evidence here needs to be stated precisely rather than gestured at, because loose claims about "billionaires owning everything" do not survive scrutiny and the Foundation states plainly where they fail. Institutional asset managers, not individual billionaires, hold the majority of listed U.S. equity — OECD data places institutional holdings at roughly 65% of listed U.S. equity, and most of that beneficial capital belongs to pensioners, retirement savers, insurers, and other diversified investors, not to the asset managers' own executives.7 A claim that a small billionaire class holds majority economic ownership of the automation stack is not supported by available evidence and should not appear in this paper's argument.
What the evidence does support is narrower and still consequential: economic ownership, voting control, and operational control are three different quantities that concentration can be measured on, and they do not move together. As of company filings in 2025–2026: Alphabet's founders held roughly 52.7% of voting power at the end of 2025 through a dual-class share structure, despite holding a smaller share of total economic equity;8 Meta's founder held roughly 60.8% of voting power as of April 2026 while holding roughly 13–14% of economic interest;9 and a single individual held a reported 19.9% economic stake in a major vehicle, AI, energy, and robotics firm as of a July 2026 filing — short of majority ownership, but an extraordinary concentration of position in a firm spanning several of the domains this paper's taxonomy treats as high-essentiality.10 The general pattern these examples illustrate: a founder does not need majority economic ownership to control board composition, strategic direction, model deployment policy, data governance, acquisition decisions, and infrastructure investment. Separately, the fifty largest institutional investors held on the order of $20 trillion in listed equities as of 2022, with the five largest holding roughly $8 trillion between them — a second, distinct concentration layer sitting alongside founder voting control, and one the OECD itself treats as a live competition-policy concern through its work on common institutional ownership.11
The corrected proposition, stated at the level of evidence rather than intuition: ownership and effective control over frontier AI, industrial robotics, cloud infrastructure, communications platforms, and increasingly defense technologies are already concentrated among a relatively small group of founders, financial institutions, corporations, and states — not through majority economic ownership by any single class, but through the layered combination of voting control, operational control, and institutional asset concentration described above. Automation could increase the productive returns and political leverage attached to those existing positions unless ownership and governance authority diffuse at a comparable pace to productive capacity.Level C
One structural implication of automation for this systemic question deserves stating even though it cannot yet be formally measured: traditional coercive and productive systems have historically required the cooperation of large numbers of people — workers who could withhold labor, organize, leak information, or defect. That dependency has functioned as an informal check on concentrated power independent of any law or constitution. Automation does not eliminate that check by reducing headcount to zero; it reduces the minimum coalition required to exercise coercive or productive capacity, which is a different and more specific claim than "fewer jobs." A smaller required coalition is, mechanically, an easier one to keep aligned, informed, or coerced into compliance than a larger one — the Foundation flags this as among the more important open questions this paper does not yet resolve, and returns to it directly in Section XII's treatment of autonomous enforcement.
This section is retained from the paper's draft phase and lightly revised; the mid-document reference list and revision log previously here have been consolidated into the full reference list and closing notes at the end of the paper.
The data-dividend proposal begins from a real observation: contemporary AI systems derive substantial economic value from data people generate, often without direct compensation. Posner and Weyl's "data as labor" argument, developed at length in Radical Markets (2018), is the serious academic origin of this idea — not, as informal commentary on this topic sometimes implies, a novel proposal invented in the course of casual discussion.12 Under a data-dividend system, firms would owe individuals or collective data trusts compensation when their data contributes to commercially valuable models.
Given the attribution problem, a data dividend is more plausible at the collective rather than individual level: sector-wide data royalties, data-trust licensing, statutory levies on data-intensive commercial AI, or negotiated group rights and provenance-based licensing for identifiable copyrighted or professional material — mechanisms that sidestep the need to compute any single person's marginal contribution by instead pricing data access at the pool or category level. Even so, the Foundation does not regard a data dividend as capable of becoming a universal, economy-wide replacement for wage income: some individuals and activities generate commercially valuable data; many do not, and a rights-based distribution system built to address the participation problem RP04 and this paper describe should not make basic participation contingent on the market value of one's behavioral exhaust, which would replicate exactly the "conditional, revocable claim" problem Section III raises about programmable currency, in a different guise. A data dividend is best understood as a supplemental term in RP04's identity — most plausibly folded into Ki or Ri, depending on design — not as a freestanding solution.Level C
A more general mechanism than a data dividend distributes a portion of measured automation-attributable productivity gains directly — RP04's automation dividend (Section 9.3) and robot productivity dividend (Section 9.7) mechanisms, examined here for their governance rather than funding properties.
A more administratively tractable path, and the one RP04 Section 9.7 gestures toward without fully developing, taxes outcomes that are already disclosed rather than attempting to isolate automation's specific causal share: excess corporate cash flow, monopoly profit, capital gains, land value, automated-system licensing revenue, or compute usage above a defined threshold. This sacrifices precision (the tax base is a proxy for automation gains, not automation gains themselves) for administrability (the underlying figures are already reported under existing disclosure regimes, as RP04 Section IV's Amazon case study demonstrates), and the Foundation regards that tradeoff as more honestly stated than pretending a clean one-to-one mapping between a specific machine and a specific former job is achievable at policy scale.
Regulatory capture — the tendency of regulatory bodies to be influenced, and sometimes controlled, by the industries they regulate — is a serious, decades-old field within political economy, not a partisan accusation. George Stigler's foundational 1971 formulation argued that regulation is, in significant part, acquired by industry and designed and operated primarily for its benefit;13 the broader public-choice literature associated with James Buchanan extends this into a general account of how concentrated interests systematically outorganize diffuse ones in shaping policy.14 The Foundation states this literature's actual claim precisely because informal discussion of this topic tends to compress it into "regulation is always captured," which the literature itself does not support — capture is a variable matter of degree, contested empirically case by case, not an all-or-nothing condition.
Countervailing evidence belongs in the same paragraph as the concern, not in a separate rebuttal section: courts have overturned agency actions industry favored; privacy legislation has passed over sustained industry opposition in multiple jurisdictions; antitrust actions, including the FTC's 2023 suit against Amazon, have proceeded against well-resourced defendants; new entrants continue to erode incumbent positions in technology markets faster than in most other capital-intensive sectors; and federalism and jurisdictional competition give affected parties routes around a captured regulator that a unitary system would not. Capture is a real risk requiring vigilance, not a settled description of how regulation universally functions, and the Foundation treats claims in either direction — "regulation always works" or "regulation is always captured" — as failing the same evidentiary standard this paper applies elsewhere.
A more precise version of the "technology moves faster than law" intuition is measurable rather than merely felt. The Foundation proposes Governance Response Latency (GRL) as a candidate construct:
A claim adjacent to this paper's argument, common in informal discussion of automation and abundance, holds that AI drives the marginal cost of production toward zero across the economy, ending scarcity as an organizing economic condition. This overstates the case. Automation can reduce marginal cost in specific, task-substitutable domains; it does not eliminate scarcity in land, energy at particular times and locations, compute, minerals, desirable housing, healthcare attention, physical space, or ecological capacity — every highly automated economy this paper's scenario analysis (Section XIV) considers still confronts genuine scarcity in at least these categories. The more defensible formulation, and the one this paper adopts: industrial capitalism developed under conditions in which productive output remained strongly coupled to human labor, while AI and robotics may weaken that coupling across a widening range of activities without eliminating scarcity as such. A distinct and separable phenomenon is ownership-based exclusion — scarcity that is not physical but institutional, arising because access to an abundant or near-zero-marginal-cost resource is nonetheless restricted by ownership, licensing, or platform terms rather than by physical limits. Distinguishing genuine scarcity from ownership-based exclusion matters directly for this paper's argument: Section III's programmable-currency risk and Section VII's ICRI are both, at root, concerned with ownership-based exclusion of an otherwise-abundant claim, not with genuine physical scarcity, and conflating the two — treating institutional exclusion as though it were an unavoidable resource constraint — would misdiagnose exactly the risk this paper is built to identify.Level B
A common informal argument holds that concentrated wealth-holders are "betting on a world that won't exist" — that automating away the wage-earning population who would otherwise be their customers is self-defeating. Section II's claim identity already shows why this does not survive as a factual claim: capital-derived claims (Ki) remain fully meaningful under automation and may concentrate further rather than lose value, because ownership of automated productive capacity is a claim on real output regardless of whether that output is purchased primarily by a broad wage-earning consumer base or by a narrower set of institutional, governmental, and high-wealth buyers. A sufficiently automated economy could, in principle, remain profitable while serving a much narrower customer base — governments, other firms, wealthy households, and military institutions — inefficiently, unequally, and with reduced aggregate output relative to its productive potential, but not necessarily unprofitably or in a way that collapses on any predictable timeline. The Foundation does not adopt the stronger claim that concentration is "self-defeating" or represents an "evolutionary dead end" — durable, highly unequal economic systems have persisted historically for long periods, and there is no basis for assuming domination resolves itself. The research-legitimate form of the underlying intuition, and the one this paper adopts: current investment incentives remain largely organized around maximizing returns within labor-mediated market structures, even as AI may progressively weaken the coupling between labor and production; whether existing ownership incentives remain stable, and whether owners' collective interest in preserving broad-based demand (real, per RP04 Section VI) outweighs each individual owner's short-run interest in automating faster than competitors (also real, per RP04 Section III's collective-action framing), is accordingly an open institutional question this paper treats empirically rather than assumes resolved in either direction.
The Foundation's second new construct, flagged since Section VII as ICRI's necessary complement.
The degree to which a person or community can maintain essential economic participation after leaving, or being excluded from, a dominant platform, issuer, jurisdiction, or infrastructure provider.
Per Section 7.6: ICRI measures concentrated capability and discretion held by institutions; EEC measures the alternatives and protections available to the people that capability could be exercised against. An institution can score high on ICRI while affected individuals retain moderate EEC because genuine alternatives exist elsewhere in the system — high convergence risk with a working escape route. Conversely, a modestly integrated institution can create severe harm where affected people have no practical alternative — low convergence risk that is nonetheless locally devastating because EEC is near zero. Neither construct alone captures the paper's central concern; the provisional combined measure, Convergence Exposure Risk (CER = ICRI × (1 − EEC), both normalized), remains flagged rather than formalized until EEC itself has undergone the same validation scrutiny ICRI received in Section VII.
Autonomous and semi-autonomous systems relevant to enforcement — drones, perimeter-security robots, predictive-policing tools, facial-recognition systems, automated threat classification, and remotely-operated or autonomous weapons — warrant direct attention without being converted into unsupported prophecy about robotic armies. The governance question live in 2026 is not hypothetical: a bipartisan trio of U.S. representatives introduced the Human Authority over Autonomous Weapons Act in July 2026, requiring human oversight, approval, or a human-in-the-loop protocol for any intentionally lethal use of an autonomous or AI-enabled weapon system, and mandating non-AI verification of AI-generated targets for five years following enactment;15 separately, a Senate Armed Services Committee proposal the same month recommended a regulatory framework emphasizing human judgment and ultimate human responsibility over military AI and autonomous weapons.16 That lawmakers from both parties regard machine-directed force as a present governance question, not a remote one, is itself evidence for treating the domain seriously rather than dismissing it as speculative fiction.
The risk this paper's framework identifies is not that autonomous systems acquire independent intent — Section III's non-anthropomorphization principle applies here as much as anywhere else in this research program. The risk is structural: human institutions deploying such systems under rules that reduce accountability, compress human judgment, and lower the cost of exercising coercive power at scale. Historically, coercive systems have required recruiting, training, paying, and retaining large numbers of people — police, soldiers, guards, administrators — each of whom could refuse an order, organize, leak information, defect, or simply quit, functioning as an informal, non-legal check on how coercive power could be exercised regardless of what the law formally permitted. Section 7.8 named this the minimum-coalition effect: automation does not need to eliminate human involvement in enforcement entirely to weaken this check; it only needs to shrink the number of people whose active cooperation a given exercise of coercive power actually requires.Level C
A responsible research and governance agenda in this domain, building directly on the ICRI/SICRI apparatus developed in Section VII, should examine: mandatory human authorization for any use of force, with the verification requirement the pending federal legislation already proposes; prohibited categories of fully autonomous targeting; audit logs and continuity records adequate for after-the-fact review; clear legal personhood and liability chains when an autonomous system is involved in a harmful outcome; geographic and functional restrictions on deployment; independent inspection regimes; shutdown and override requirements enforceable against the deploying institution, not merely the system itself; whistleblower protections for personnel inside enforcement-adjacent automation programs; and — the requirement this paper's governance principle (Section 7.7) already states in general form — strict separation between civilian benefit-administration databases and any enforcement or surveillance database, so that a person's economic participation claim (Section II) cannot be conditioned on, or discovered through, an enforcement-oriented data system never authorized for that purpose.
Section 5.4 evaluated community and cooperative ownership as one of four architectures; this section briefly extends that treatment in light of Sections VII through XII rather than repeating it. Community-scale infrastructure — cooperative energy grids, municipal compute, cooperative agricultural robotics, local fabrication systems, resident-owned housing-production systems, regional data trusts — is not primarily valuable, on this paper's analysis, as a route to matching frontier-scale private or state capacity. Its distinct value is as a structural floor: a form of Economic Exit Capacity (Section XI) that exists independent of whether any dominant national or global provider remains trustworthy. A community that retains even modest independent energy, food, compute, or fabrication capacity holds a nonzero Aj term in the EEC equation regardless of what happens to ICRI at the national or systemic level — which is a different and, in the Foundation's judgment, more defensible justification for investing in community-scale productive capacity than treating it as a comprehensive alternative to concentrated ownership, which Section 5.4 already found it is not currently positioned to be.
Ordinal, qualitative scoring throughout — consistent with this paper's practice elsewhere of not manufacturing false quantitative precision where none currently exists.
| Scenario | ICRI Trajectory | EEC Trajectory | Institutional Legitimacy | Implementation Complexity |
|---|---|---|---|---|
| A — Concentrated Automation, No Reform Automation gains accrue primarily to existing owners; RP04's Section IX "no reform" case. | Rising — Section 7.8's evidence suggests existing voting/operational concentration compounds rather than dilutes absent intervention | Falling — fewer independent alternatives as concentrated providers absorb adjacent domains | Falling, per Section I's methodology note — perceived unfairness compounds even where formal legality is unchanged | Lowest — the status quo requires no new institution-building |
| B — Universal Cash Distribution via Existing Currency RP04 mechanisms funded and paid through ordinary dollars, no new token. | Roughly flat — does not directly address ownership or convergence | Roughly flat — preserves existing alternatives, adds none | Depends heavily on funding-source legitimacy (Section 10.1's capture question) | Low — RP04 Section III's stimulus-payment precedent shows the plumbing exists |
| C — Programmable Consumption Credits With or without demurrage, per Section III–IV. | Rising, absent the safeguards Section III specifies — programmability concentrates revocation control (Section I) in the issuer | Falling, unless an explicit non-programmable fallback is preserved | Contested — Section III's own evidence (the 2026 CBDC legislation) shows this exact tradeoff is a live, high-salience legitimacy question | Moderate — technology is mature (Section III), governance safeguards are not |
| D — Universal Capital Ownership Section 5.3's citizen-equity mechanism, non-transferable core. | Falling, if genuinely broad-based and the non-transferability safeguard holds; Section 5.3's testable hypothesis is directly relevant here | Rising — a durable capital claim is itself an EEC input (Section XI) | Plausibly rising — addresses the "who owns automation" question this paper is organized around directly | High — capitalization source (Section 5.3) and legal-instrument design remain open |
| E — The Candidate Mixed Plural Architecture Section VI's H1, tested against H2/H3. | Falling under H1; unresolved under H2/H3 — the scenario's actual outcome is an empirical question this paper does not claim to have answered | Rising under H1, via redundant/plural alternatives | Plausibly highest if H1 holds — plural, contestable governance is itself the design goal | Highest — requires coordinating private, public, citizen, and community institutions simultaneously |
Offered as research principles for further scrutiny and refinement, not a finalized constitutional instrument or a Foundation policy position.
This paper's governance implications are institutional rather than technological, consistent with RP04's own conclusion. If Sections II through XII are directionally correct, the appropriate response is not primarily AI capability regulation — a domain the Foundation's other publications address — but building the measurement infrastructure this paper proposes (ICRI, SICRI, EEC, and RP04's LRT/EPI/PE/DRR), the institutional separations Section VII's governance principle specifies, and the political capacity to apply them before convergence compounds past the point where ordinary competition law, financial regulation, or civil-rights enforcement can address it after the fact. A jurisdiction currently has no standardized indicator comparable to ICRI for detecting cross-domain institutional convergence before it becomes politically consequential, and no established mechanism connecting automation-driven ownership concentration (Section VII, X) to the participation-restoration mechanisms RP04 proposes. Building both — the measurement and the mechanism — is the governance gap this paper's research program is oriented toward closing, independent of which specific ownership architecture (Section V, VI) or funding mechanism (RP04 Section IX, this paper's Sections VIII–IX) ultimately proves tractable in any given jurisdiction.
As with RP04, the Foundation states these conditions because a thesis that cannot specify how it could be wrong cannot be meaningfully debated — only agreed with or dismissed.
RP04 argued that employment has functioned as capitalism's distribution mechanism, not its purpose, and that automation threatens the mechanism rather than the purpose. This paper has argued that even a fully repaired distribution mechanism — an automation dividend, a sovereign fund, universal citizen equity, any of RP04's eleven mechanisms — does not by itself answer who governs the resulting claims, on what terms, and with what recourse. A currency or credit system can solve RP04's demand-side arithmetic while still being an instrument of conditional permission rather than genuine participation.
Robots do not decide whether their output becomes concentrated profit, public wealth, or broadly held capital. Tokens do not decide whether they function as freedom or as a leash. Those outcomes emerge from the institutions this paper has tried to make evaluable rather than rhetorical: the Claim Governance Lifecycle a distribution mechanism is subject to, the ownership architecture that controls productive capacity, the degree of institutional and systemic convergence across the domains a person's economic life depends on, and the exit capacity available if any part of that system fails or turns against them. None of this paper's evidence supports the claim that a coercive outcome is intended, planned, or inevitable. Section 7.8's evidence does support a narrower and still serious claim: existing ownership, voting, and institutional concentration in the automation stack is not starting from a neutral baseline, and diffusing the governance authority automation is generating is not a task markets accomplish automatically as a byproduct of growth.
The question this paper has been organized around, restated plainly: not whether automated productive capacity will exist — it already does, and RP04's evidence suggests its scope will grow — but whether the claims that capacity generates remain governed by institutions people can meaningfully appeal to, exit, and trust, or whether they become conditional permissions administered by systems no external check can reach. That outcome is not technologically predetermined. It is, as this paper has tried to state in evaluable rather than rhetorical terms throughout, a design choice — one this paper's proposed measurement instruments are built to help make visible before it is made by default.
| Symbol | Definition | First Used |
|---|---|---|
| Qi | Total recognized economic claim held by person i | Section II |
| Wi, Ki, Ti, Ri, Di | Wage, capital, conditional-transfer, rights-based, and debt-financed components of Qi | Section II |
| B(t), λ | Demurrage-subject balance and its decay rate | Section IV |
| xi, wi, aij, dij, sij, α | ICRI domain control, risk weight, integration, discretion, safeguard, and balancing parameter | Section VII |
| ci, ei, ri | ICRI domain control's dependency weighting: share of control, essentiality, substitutability | Section 7.2 |
| Co, Cv, Ci, Cd, Cs, Ce | SICRI's six proposed concentration inputs | Section 7.8 |
| Aj, vj, Sk, ck | EEC's alternative-pathway and switching-cost terms | Section XI |
| GRL, TDT, Γ | Governance Response Latency, Technology Diffusion Time, and their ratio | Section 10.2 |
This paper extends, and does not restate, RP04's core apparatus. RP04's identity, Pt = Wt + Ct + Tt + Dt, is disaggregated to the person level as Qi in Section II, with Tt split into conditional transfers (Ti) and rights-based distributions (Ri). RP04's LRT, EPI, PE, and DRR constructs remain this paper's companions rather than being superseded; ICRI, SICRI, and EEC extend the panel from measuring distribution and participation to measuring governance and convergence. RP04's eleven participation-restoration mechanisms (its Section IX) are the raw material for this paper's Sections VIII, IX, and V–VI, examined here for governance properties RP04's funding-and-tractability framing did not address. RP04's Participation Chain (Production → Distribution → Participation → Demand → Economy) and this paper's Claim Governance Lifecycle (Creation → Allocation → Recognition → Exercise → Conversion → Transfer → Modification → Revocation → Appeal → Exit) are offered as complementary organizing frameworks at different levels of granularity — the former describing the macroeconomic sequence, the latter describing what happens to a single claim within it.