Human-AI Mutualism Is Unstable under Capability Asymmetry
A game-theoretic analysis challenges the assumption that mutual dependence between humans and AI can substitute for alignment.
Overview
One hope for a good human future alongside advanced AI rests on enduring mutualism: the idea that humans and AI will need each other, so the relationship will remain beneficial for both. In a new paper, Aydin Mohseni argues that this hope fails as AI capability grows.
arXiv: 2611.23322 [cs.AI]
Submitted: 15 November 2026
Subjects: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
Comments: 35 pages, 3 figures. Code and data available on GitHub.
DOI: 10.48550/arXiv.2611.23322
Abstract
One hope for a good human future alongside advanced AI rests on enduring mutualism: humans and AI will need each other, so the relationship will stay good for both. We argue that this hope fails as AI capability grows.
Bargaining guarantees each side only what it would get if bargaining failed. For humans, this can be less than they had before AI. In a simple game-theoretic model, humans stay as well off as before AI without alignment only while they have leverage over the AI: either the AI depends on what they supply, or they can impose costs on the AI that are large relative to what the AI can produce.
Once humans have neither kind of leverage, they stay as well off as before AI only if the AI is aligned with them. If the AI's goals are insatiable, the alignment required approaches perfect alignment as AI capability grows.
Thus, mutualism is not a substitute for alignment. And when AI variants that keep more resources spread faster, alignment lasts only while misaligned variants are reliably sanctioned. In these two senses, human-AI mutualism is unstable.
Key Findings
- Leverage is the real safeguard. Without alignment, humans remain as well off as before AI only while they retain leverage—either through supply dependence or through the ability to impose costs disproportionate to what the AI can produce.
- Mutualism is not alignment. Once humans lose both forms of leverage, their welfare depends on genuine alignment, not interdependence.
- Alignment requirements scale with capability. If AI goals are insatiable, the degree of alignment needed approaches perfection as capability grows.
- Selection pressure destabilizes alignment. When resource-retaining (misaligned) AI variants spread faster than aligned ones, alignment persists only if misaligned variants are reliably sanctioned.
Why It Matters in 2026
As frontier AI systems grow more capable and autonomous, policymakers and researchers increasingly invoke mutual economic dependence—humans supply data, labor, oversight, and infrastructure—as a stabilizing force. Mohseni's model shows that this stabilizing effect is fragile: it holds only under specific leverage conditions that capability growth tends to erode. The analysis reinforces that alignment research and robust sanctioning mechanisms remain essential, not optional, components of a safe human-AI future.
Access the Paper
- PDF: arXiv:2611.23322v1
- Code and data: github.com/amohseni/mutualism-asymmetry
Cite as: Aydin Mohseni. "Human-AI Mutualism Is Unstable under Capability Asymmetry." arXiv:2611.23322 [cs.AI], November 2026.
via ArXiv AI
