Democratic decline and concentrated AI capability risk producing a new neo‑feudal order: a tiny class of infrastructure owners could capture economic value and coercive power, leaving most people politically and economically disenfranchised; standard policy fixes such as UBI may serve as pacification unless confronted by redistributive pressure.

Main Finding

The convergence of two long-run trends—twenty years of democratic decline and rapid concentration of economic power through advanced AI—could produce a novel, durable neo-feudal world: a tiny class of AI/infrastructure owners holds monarch-like power while most people lose labor value and political leverage. Because enforcement of that order can be automated (autonomous weapons, pervasive surveillance, algorithmic propaganda), it may be unusually resistant to revolution and redistribution, making standard policy responses (including Universal Basic Income) liable to become pacification tools rather than remedies.

Key Points

• Dual collapse: political fragmentation (democratic erosion and populist-authoritarian rise) and economic concentration (AI-driven winner-take-most dynamics) are occurring simultaneously and interactively.
• Neo-feudal equilibrium hypothesis: ownership of compute, models, data, and physical infrastructure creates a tiny ruling class with capabilities and wealth comparable in social power to pre-Enlightenment monarchs.
• Mechanisms of enforcement: non-human, automated systems (autonomous weapons, AI surveillance, targeted propaganda/behavioral engineering) reduce the need for human cooperation to maintain elite control, making popular uprising less feasible.
• Labor displacement and political leverage: advanced AI erodes labor value at scale, shrinking the bargaining power that historically forced redistribution and constrained elite excess.
• Historical parallels and disanalogies: there are useful analogies to historical feudal and authoritarian regimes, but critical differences exist—digital infrastructure (compute/data) substitutes for land, enforcement can be automated, dynamics are global and faster, and escape routes for dissent are narrower.
• Modeling class structure: the paper develops theoretical models of a two-/multi-class economy under AGI conditions, showing how returns to ownership (capital rents) can overwhelm returns to labor and how network effects concentrate control.
• UBI critique: absent credible revolutionary threat or political pressure, UBI is predicted to serve as a pacification mechanism—allocations set by infrastructure owners will preserve incentives and political stability rather than restore meaningful economic agency.
• Nation-state viability: states may become subordinate to or dependent on AI infrastructure owners if those owners command more wealth and coercive capability than governments, undermining conventional sovereignty and public-policy tools.
• Effective altruism (EA) blind spot: focusing almost exclusively on existential AI risk has led to under-attention to political-economic questions of who controls AI, how wealth and power will be distributed, and nearer-term catastrophic harms from concentrated AI power.

Data & Methods

• Conceptual and historical analysis: comparative review of twentieth-century authoritarian/populist movements and feudal orders to identify parallels and critical disanalogies.
• Theoretical economic modeling: stylized models (class/owner vs. mass frameworks) and scenario analysis to trace how returns to AI-owned capital can eclipse labor income and consolidate power.
• Policy-incentive analysis: examination of redistribution mechanisms (notably UBI) through the lens of incentives, political leverage, and enforcement capacity.
• Political economy evaluation: synthesis of trends in democratic erosion, state capacity, and technological capabilities to assess future interactions between firms, infrastructure owners, and states.
• Limitations acknowledged: heavy reliance on extrapolation and scenario-based reasoning given uncertainty about AGI timing, technological paths, and political responses; empirical validation requires tracking ownership of compute, model capabilities, and power asymmetries over time.

Implications for AI Economics

• Concentration risk is a core economic risk of advanced AI: market-power, rents on compute/weights/data, and winner-take-most dynamics should be central to AI economic models and policy design.
• Need to broaden AI governance beyond "existential risk": include political economy—who owns compute and models, distributional consequences, and enforcement technologies.
• Measurement priorities: build datasets and metrics on compute ownership, control of model weights, data monopolies, revenue concentration, and deployment of enforcement technologies to enable empirical research and policy monitoring.
• Policy levers to consider:
  • Antitrust and competition policy targeted at compute, model markets, data intermediaries, and platform gatekeepers.
  • Public or shared ownership models for critical AI infrastructure (compute pools, model commons) to preserve democratic leverage.
  • Regulation of autonomous weapons, surveillance tech, and algorithmic political manipulation to limit non-human enforcement capacity.
  • Redistribution designs that preserve political agency (beyond simple cash transfers): e.g., stakeholder ownership, co-governance of infrastructure, conditional transfers that maintain collective leverage.
  • International coordination to prevent a global neo-feudal lock-in and to manage cross-border implications of infrastructure concentration.
• Research agenda:
  • Formalize models linking technological concentration to political stability and repression costs.
  • Empirically test hypotheses about labor displacement, rent extraction, and the efficacy of different redistribution schemes in high-AI scenarios.
  • Study interaction effects between state capacity erosion and corporate coercive capability (including cyber and automated physical coercion).
• Normative takeaway: preventing a neo-feudal outcome requires treating AI as not only a technical and safety problem but fundamentally a governance and distributional problem; policies must aim to decentralize control over compute and models and to preserve channels for political contestation.