What happened
RAND's Center for the Geopolitics of Artificial General Intelligence published this 40-page Expert Insights paper arguing that global AI competition 'should be understood not just as a sprint to develop the most-advanced systems but also as a longer-term marathon to accumulate net benefits — the gains from AI minus the costs of accidents, misuse, and disruption.' Drawing on Charles Perrow's theory of 'normal accidents' in complex technological systems, author Brian Jackson models four intermediate scenarios showing how differing levels of AI-risk cost affect national and global net benefits, concluding that 'even modest resilience advantages could compound into decisive competitive edges over time' — potentially outperforming nations that won the initial capability race. The paper closes with concrete policy options for individuals, firms, infrastructure operators, and state/local/national governments to build resilience against recurring AI-related disruption.
Why it matters
Reframes national AI competitiveness debates around risk-adjusted net benefit rather than raw capability leadership, giving policy and industrial-strategy audiences a fresh analytical lens — and a resilience-investment case — to bring into AI strategy and national-security discussions.
Action needed
Incorporate the net-benefit/resilience framing into national or corporate AI strategy briefings and evaluate current AI risk-management posture against the paper's four intermediate-scenario model.