AI & Nuclear Stability | Crisis Decision-Making Research

AI and the Future of Nuclear Stability & Crisis Decision-Making

This project examines how generative AI and AI-enabled decision-support systems are reshaping nuclear crisis decision-making and strategic stability. As governments explore AI tools to simulate crises, interpret adversary behavior, and support high-stakes decisions, these systems promise enhanced foresight but also introduce risks of misperception, automation bias, and inadvertent escalation.

Supported by Longview Philanthropy and the Carnegie Corporation of New York-led Consortium to Reduce Nuclear Dangers, the project develops policy-relevant frameworks to ensure human judgment remains central in AI-mediated crises and to reduce the risk of miscalculation or inadvertent escalation.

Focus

How AI reshapes perception, signaling, and escalation dynamics
Cognitive and ethical risks in human–AI decision-making
Governance and oversight for responsible AI integration
Preserving human agency in high-stakes strategic environments

Contribution

Synthetic deterrence: AI-generated simulations reshape coercion and credibility
Human–machine cognition: decision-makers’ cognitive styles shape how AI outputs are interpreted and trusted
Ethics and accountability: a hybrid framework to preserve human moral responsibility
AI as an epistemic actor: intelligent systems shape perception and strategic belief, not merely analysis

Core Output

The project’s principal scholarly output is Machines at the Brink: AI and the Future of Strategic Crisis Decision-Making (under contract with Oxford University Press). The book synthesizes the project’s theoretical, empirical, and policy insights and forms part of a broader research arc with AI and the Bomb and The AI Commander, examining how intelligent systems are transforming nuclear strategy, military decision-making, and strategic stability. Together, the project and book aim to strengthen institutional resilience and provide a roadmap for managing strategic stability in an era where intelligent systems increasingly mediate crisis decision-making.