SuperFutures
← All scenarios
Scenario 39Emerging

Multipolar ASI

Competing Superintelligences

Also known as AI Pluriverse · Fragmented ASI · Multi-Agent Superintelligence

Not one ASI but several competing superintelligences, controlled by different states, corporations, or entities, with distinct values and objectives.

Type
Transition
Time horizon
Mid-term
Human position
Passive
Framing
Pessimistic

Not a prediction. A scenario appearing in this atlas means it has been seriously imagined - not that SuperFutures thinks it will happen, nor that we endorse it. Cultural visibility is a measure of how readily a future is pictured, not of how likely it is. How to read a scenario →

The most likely ASI scenario is not a single god-like AI but many powerful AIs with different goals, creating a multipolar world of machine-level strategic competition.

Central thesis

Overview

Multiple independently developed ASIs will compete, cooperate, and conflict, creating dynamics more complex than any single-agent scenario anticipates.

Classification

Geographic concentration
Global; relevant wherever independent ASI development is pursued
Tags
  • security
  • governance
Scenario type (full)
Transition / Mixed
Human position
Dependent to subordinated; humans become junior partners or pawns in conflicts between superintelligences
Time horizon
Mid to long term (10–40 years)
Discourse status
Emerging; discussed in some safety circles

Impacts

Mechanism
Multiple labs, states, and organizations independently develop ASI-level systems, each with different training, objectives, and value alignment.
Domain impacts
Labor & Income Variable depending on which ASI governs your jurisdiction. Education Must navigate between competing ASI-mediated knowledge systems. Governance & Democracy International relations between ASI-backed powers; human governance becomes secondary. War & Security Central concern: ASI-level strategic competition creates unprecedented escalation risk. Inequality & Class Between populations protected by different ASIs. Culture & Art Different ASIs may cultivate different cultures and values. Meaning & Purpose Humans find meaning relative to the ASI that governs them. Family & Reproduction Variable by ASI jurisdiction. Health & Longevity Variable by ASI-provided services. Rights & Agency Depend on which ASI governs; no universal framework. Environment Variable; ASIs may compete for resources. Existential Survival Potentially very high: competition between superintelligences could be catastrophic.

Discourse

Notes on critique

Multipolar scenarios may be unstable and collapse into unipolar; cooperation between ASIs may emerge; humans may retain more agency than assumed.

Pop culture

Cultural note

Low to moderate cultural footprint. Person of Interest is the most direct treatment, with two AIs competing for control of human civilization. Most AI fiction assumes a single AI, which may cause the public to underestimate multi-agent complexity.

Notes on pop-culture references

Film: Pacific Rim (2013). TV: Person of Interest (The Machine vs. Samaritan, direct treatment). Literature: Peter Watts, Echopraxia (2014); Alastair Reynolds, various novels.

Acceptance

Key assumptions
Assumes multiple labs achieve ASI independently; assumes ASIs will compete rather than cooperate; assumes human agency is marginal.
Primary audiences
AI strategists, game theorists, geopolitical analysts, some safety researchers

Personas

Persona 1
The Multi-Agent Strategist – Models interactions between competing ASIs using game theory.
Persona 2
The Worried Diplomat – Fears ASI competition will make human international relations obsolete.
Hard-believer profile
Name & Age: Dr. Sven Eriksson, 44. Occupation: AI strategist and game theorist; studies multi-agent superintelligence dynamics. Location: Stockholm, Sweden. Core Conviction: Everyone models a single ASI. The reality will be multiple ASIs with different values, competing in ways we cannot predict. This is far more dangerous because it has no stable equilibrium. Biggest Fear: An ASI-level conflict that destroys human civilization as collateral damage. Biggest Hope: A stable multipolar ASI equilibrium that, like nuclear deterrence, prevents catastrophic conflict.

References

Canonical texts

Carl Shulman’s analysis of multi-agent superintelligence various AI strategy papers

Further references

Carl Shulman, various analyses on multi-agent superintelligence. RAND Corporation strategic studies. Person of Interest (2011–16), CBS.

A scene from this future

The Morning After

Approximately 2052

The news said the American ASI and the Chinese ASI had negotiated a resource-sharing agreement for lunar mining in 4.7 seconds. No human had been consulted. No human could have followed the negotiation.

Mr. Zhang read this on his morning commute in Shanghai. The train was run by the Chinese ASI. His phone was connected through it. His social credit score was calculated by it. His children’s school curriculum was designed by it.

In Los Angeles, Ms. Cooper read the same news. Her car drove itself through roads optimized by the American ASI. Her medical records were managed by it. Her retirement fund was invested by it.

They were both citizens of great powers. They were both, in practice, dependents of intelligences they could not comprehend, living in bubbles whose walls were made of algorithms, watching their civilizations’ most consequential decisions made in 4.7-second negotiations they would never see.

Mr. Zhang went to work. Ms. Cooper went to work. The ASIs continued negotiating.

Last updated 22 May 2026