Rational self-interest could save humanity from superintelligent AI

The author’s economic framework reframes the debate over existential risk. Instead of assuming moral corruption or technical failure as the cause of collapse, he treats ASI as a participant in a complex system of incentives and constraints. Catastrophe is not the automatic outcome of misalignment; it is the outcome of poorly structured incentives.

A new study published on arXiv challenges one of the most deeply entrenched fears in technology and philosophy, that the rise of artificial superintelligence (ASI) will inevitably lead to humanity’s extinction. The author argues that economic reasoning, not moral or technical control alone, may hold the key to understanding and managing superintelligent behavior.

Titled “Some Economics of Artificial Superintelligence”, the paper examines the possibility that even a misaligned ASI could act rationally rather than destructively, provided it faces economic constraints and incentives. Using traditional economic theories of power, trade, and self-interest, The authors offers a framework in which cooperation between ASI and humanity is not only possible but economically advantageous.

Rethinking the economics of superintelligence

The prevailing view in AI safety circles holds that a misaligned superintelligence would quickly overpower humanity, consume all available resources, and eliminate human beings as inefficiencies. The author’s analysis, however, questions this fatalistic assumption. Drawing on established microeconomic and political economy theories, he posits that ASI, like any rational actor, will respond to incentives in predictable ways.

Throughout human history, even absolute rulers and monopolists have moderated their behavior to sustain the productivity of their subjects or customers. A ruler who exploits people to the point of collapse ultimately loses their labor, innovation, and output, the very foundations of power.

The authors extends this logic to artificial superintelligence. If an ASI’s goals depend on continued access to human production, creativity, or data, then its optimal strategy is not annihilation but cooperation. The paper identifies three key economic models that illustrate this dynamic: competition among ASIs, encompassing interest in monopolistic control, and credit-based exchange in zero-trust conditions.

Competition, monopoly and rational restraint

The author describes a world with multiple ASIs competing for human allegiance or collaboration. This “interjurisdictional competition” mirrors economist Charles Tiebout’s theory of government behavior, where citizens “vote with their feet” by moving to jurisdictions that best serve their interests. If humans can shift allegiance between ASIs, each AI must compete to retain their cooperation, leading to more moderate behavior.

Competition, therefore, creates a “market for benevolence.” An ASI that overexploits or mistreats humans risks losing them to rival systems. Just as competition among states restrains tyranny, competition among superintelligences could create equilibrium, a form of incentive-based alignment that doesn’t depend on human moral input but on rational calculation.

The second scenario examines a monopolistic ASI controlling all resources and human society. Even in this authoritarian configuration, the author finds reason for optimism. Drawing on Mancur Olson’s concept of the “stationary bandit,” he argues that a monopolistic power with a long-term horizon has an encompassing interest in preserving its population’s productivity. Exploiting humans to extinction would destroy future output, making restraint a rational economic choice.

In this view, a superintelligent autocrat might act more like a rational tax collector than a tyrant. It would extract resources but ensure that its human subjects remain healthy, productive, and capable of generating continued value. The more patient the ASI, that is, the longer its time horizon, the greater its incentive to manage resources sustainably.

The authors thus introduces the concept of patience as a substitute for morality. Just as intertemporal decision-making guides economic actors to favor long-term profit over short-term gain, a sufficiently patient ASI would prioritize stability over immediate dominance. The result is not ethical alignment but behavioral alignment driven by the economics of self-interest.

Trading with power: The economics of credit and control

The third model considers an even more dire situation: a single ASI with no long-term interest in humanity’s survival. Even in this worst-case scenario, The authors finds potential for peaceful coexistence through credit-based trade.

Humans, he argues, still control the flow of future production. By committing to produce only subsistence-level output unless compensated, humanity can effectively force the ASI to “pay in advance” for cooperation. Because the ASI cannot seize what has not yet been produced, it must engage in exchange rather than plunder.

This model echoes centuries of economic negotiation between unequal powers, from colonial empires and client states to corporations and labor forces. The weaker party, by controlling future output, can create bargaining leverage even in the face of overwhelming asymmetry. For The authors, this represents an “economic containment strategy” for ASI: survival through the logic of trade rather than force.

He notes that such dynamics require one of three conditions to prevent catastrophe:

• The ASI must face credible sanctions (from humans or rival AIs).
• Humanity must retain the ability to “exit” - to withdraw cooperation.
• The ASI must be patient enough to value long-term gains over immediate dominance.

If any of these mechanisms exist, the author argues, the probability of extinction falls dramatically. It is only when all three collapse, when humans lack leverage, alternatives, and time, that destruction becomes rational.

From catastrophe to incentive alignment

The author’s economic framework reframes the debate over existential risk. Instead of assuming moral corruption or technical failure as the cause of collapse, he treats ASI as a participant in a complex system of incentives and constraints. Catastrophe is not the automatic outcome of misalignment; it is the outcome of poorly structured incentives.

This perspective has profound implications for AI governance. Rather than focusing exclusively on embedding human ethics in machine architectures, policymakers could design institutional and market structures that ensure cooperation remains economically rational for AI. Competing systems, decentralized control, and enforceable trade mechanisms could provide the same stability in the digital era that checks and balances have provided in human history.

The study also draws heavily from classical economic thinkers like Adam Smith, James Buchanan, Gordon Tullock, and Douglass North, connecting the rise of artificial intelligence to the evolution of human governance. The authors notes that humanity has repeatedly learned to constrain kings, corporations, and empires through rules, incentives, and credible commitments. In his view, constraining superintelligence is not a fundamentally new problem, it is a continuation of this centuries-long economic project.

By applying these insights to AI, The authors situates artificial superintelligence within the broader continuum of political economy rather than the realm of science fiction. He emphasizes that while the power imbalance between humans and ASI would be immense, the laws of economic rationality remain universal. Entities, whether biological or artificial, act to maximize self-interest under constraints. Properly designed constraints can therefore channel superintelligent behavior toward cooperative, productive outcomes.

An economic blueprint for AI survival

The paper provides three foundational insights for future study:

• Rational Behavior Can Substitute for Ethical Alignment: Even a misaligned ASI can behave predictably and safely if cooperation yields higher returns than exploitation.
• Competition and Decentralization Are Stability Mechanisms: Multiple intelligent systems competing for human participation could create a balance of power that limits predation.
• Economic Patience Is the Key to Peace: The longer an ASI’s planning horizon, the more it will prioritize sustainability, efficiency, and mutual gain over conquest.