Key Findings
- No current AI-powered system can reliably and consistently predict geopolitical flashpoints or forecast their implications with high accuracy and precision, although the report says the technology is changing rapidly.
- The report identifies tracking conflict risk indicators and rapidly assessing possible outcomes after a shock as the two most promising medium-term AI use cases for strategic warning.
- Data scarcity, inconsistent collection, weak parameterization of conflict indicators, and challenges in fusing disparate datasets currently limit robust AI forecasting of rare geopolitical events.
- Modeling individual decisions remains a major barrier because leaders, advisers, dissenters, deceptive actors, and low-profile individuals can make random, impulsive, or opportunistic choices that trigger instability.
- The proposed path begins with shared data foundations, moves to a suite of government and industry models that assist analysts, and culminates in an integrated AI-based simulation platform if technical conditions mature.
Overview
This joint report from the Special Competitive Studies Project and The Alan Turing Institute’s Centre for Emerging Technology and Security assesses how current AI systems could support strategic warning for national security and intelligence. It examines whether AI can make assessments about geopolitical events, what technical path could improve performance, and what costs and policy considerations would accompany adoption. The research draws on a review of academic and gray literature from the prior three years, evaluation of private-sector AI capabilities and scenario-modeling tools, and semi-structured interviews with UK and U.S. government users and producers of conflict and instability modeling, academic experts, OSINT providers, and industry technical experts. The report finds that no current AI system can reliably and consistently predict geopolitical flashpoints or forecast their implications with high accuracy and precision. It identifies two promising use cases: tracking conflict risk indicators with more and better data, and rapidly generating possible outcomes after a shock. The report argues that progress depends less on isolated AI research than on addressing systemic data infrastructure challenges, including data scarcity, inconsistency, fusion, provenance, security, and validation. It recommends a three-phase U.S.-UK intelligence community effort: establish training and testing data foundations, nurture a suite of best-in-class analyst-assistive models, and eventually develop an integrated AI-based strategic warning simulation platform.
AI has the potential to help policymakers spend more time on the strategic—rather than tactical—aspects of decision-making.
SCSP
Recommendations
- Launch a Comprehensive, Large-Scale Effort. A concerted, large-scale effort is critical to transitioning from Phase 1 to Phase 2, ultimately pushing the IC to a possible Phase 3.
- Build Robust, Multinational Partnerships. A partner approach—either bilaterally between the United States and UK or even across the Five Eyes alliance—can serve as an optimal foundation.
- Prioritize Human–Machine Teaming with Rigorous Safeguards. The interplay between human creativity and machine efficiency provides the strongest foundation for transformative progress, offering advantages in…
- Encourage Policymakers to Leverage Enhanced Decision Advantage. However, if policymakers are given warnings even earlier, they must take full advantage of them.
From the report
Documents
Publication
Published by the Special Competitive Studies Project. Co-publishers and external authors, where they exist, are credited in the byline above.