Advances in AI could concentrate wealth and coercive power in the hands of a tiny infrastructure-owning elite, creating a resilient neo-feudal order; autonomous weapons, pervasive surveillance and algorithmic propaganda may make such a settlement unusually hard to overturn, and policies like UBI risk becoming pacification rather than redistribution without threat-driven pressure to share power.

Main Finding

The paper argues that simultaneous geopolitical fragmentation (two decades of democratic decline) and accelerating concentration of economic power driven by AI could produce a novel, durable neo-feudal equilibrium. In that equilibrium, a tiny class of AI/infrastructure owners captures most economic and coercive power, labor value collapses, and traditional democratic checks fail because enforcement technologies (autonomous weapons, AI surveillance, algorithmic information control) can operate without human cooperation — making revolutionary overthrow far less feasible. Conventional redistribution tools such as Universal Basic Income (UBI) risk becoming pacification mechanisms rather than pathways to genuine redistribution unless they are paired with institutions that alter ownership and incentive structures.

Key Points

• Dual crises: political fragmentation + AI-driven economic concentration are mutually reinforcing and create structural conditions distinct from 20th-century authoritarian or oligarchic regimes.
• Class transformation: Advances toward general-purpose AI can generate extreme returns to owners of compute, data, and model architectures, producing a bifurcated society of infrastructure proprietors and economically (and politically) disenfranchised populations.
• Enforcement without cooperation: Autonomous weapons, pervasive surveillance, and algorithmic propaganda reduce the need for rulers to rely on human intermediaries, making repression cheaper, more precise, and harder to subvert through collective action.
• Historical analogies limited: The paper draws parallels to pre-Enlightenment feudal power concentration but stresses critical disanalogies — especially technological durability of coercion and the automation of rule — that make the new order potentially more stable and less susceptible to past revolution dynamics.
• UBI critique: UBI may be politically attractive and technically implementable, but absent redistribution of ownership or threat of expropriation, it is likely to be sized and structured as a pacification subsidy that preserves underlying property relations rather than addressing loss of labor value or political leverage.
• State erosion and privatized sovereignty: As private AI infrastructure commands more capabilities and wealth than many states, nation-state authority may wither—either becoming dependent clients of infrastructure owners or being bypassed altogether.
• Policy urgency: Remedies must focus on structural changes (ownership, access to compute, antitrust, public infrastructure, international governance) rather than one-off cash transfers.

Data & Methods

• Historical and comparative analysis: Examines twentieth-century authoritarian/populist movements and pre-Enlightenment feudal structures to identify parallels and disanalogies.
• Theoretical modeling and scenario analysis: Builds stylized models of class structure under conditions of generalized AI (e.g., returns to capital vs labor, scale economies in compute), and evaluates incentive effects of redistribution mechanisms (UBI vs ownership-based reforms).
• Institutional and political-economy reasoning: Uses mechanism-based arguments about enforcement technologies and collective-action capacity to assess stability and revolutionary feasibility.
• Normative and policy evaluation: Assesses UBI through the lens of incentive compatibility and political equilibrium, and explores the implications for state sovereignty and global governance.
• Limitations noted by the paper: largely conceptual and model-based rather than empirically validated; outcomes depend on uncertain technological trajectories (speed and form of AGI), political choices, and cross-country heterogeneity.

Implications for AI Economics

• Redistribution dynamics: If AI concentrates rents in infrastructure ownership, standard labor-market policies will be insufficient; economic models should incorporate highly skewed returns to compute and persistent capital rents.
• Measurement priorities: Economists must track concentration metrics beyond income—ownership of compute, control of training data, and access to model weights and production pipelines—to empirically evaluate capture.
• Incentive and market design: Policies that only transfer income (UBI) without changing ownership or governance of productive assets are unlikely to prevent entrenched inequality or political domination; models should analyze ownership-based interventions (public compute, mandated model-sharing, worker cooperatives, data trusts).
• Political feasibility constraints: The availability of low-cost coercion and surveillance reduces political leverage of disadvantaged groups, meaning political-economy equilibria may lock in high inequality; standard redistribution mechanisms may be politically infeasible unless international coordination or binding institutional reforms occur.
• Regulatory focus: Antitrust, limits on concentration of compute, export controls on dual-use autonomous systems, and governance of surveillance technologies become central economic-policy tools, not ancillary safety measures.
• Macro implications: Expect higher capital share of income, potential secular stagnation in labor demand, altered investment dynamics favoring compute-heavy incumbents, and possible fragmentation of global markets along infrastructure-owner lines.
• Research directions: Formal models of rent extraction by infrastructure owners; empirical work measuring compute and model concentration; policy experiments in public compute provision, worker/consumer ownership of AI outputs, and enforceable international constraints on coercive AI tools.

Overall, the paper reframes AI economics as inseparable from political-institutional arrangements: without interventions that redistribute ownership or constrain coercive uses of AI, technological progress risks producing a durable, low-mobility equilibrium with severe economic and democratic consequences.