Research Summary: Capability-Based Planning for AI Crisis Preparedness

Current government approaches to AI crisis preparedness are hampered by a 'predict-then-act' paradigm that struggles with the inherent unpredictability of AI. Experts and official reviews acknowledge that traditional likelihood-based risk assessments are unsuitable for AI-related risks. A new methodology is proposed, drawing on principles of decision-making under deep uncertainty, which focuses on capability-based planning rather than predictive risk ranking.

Why it matters

The development of robust and adaptable frameworks for AI crisis preparedness is crucial for ensuring governmental and institutional resilience in the face of rapidly evolving technological risks. Moving beyond traditional predictive models to capability-based planning offers a more practical approach to managing deep uncertainty, thereby strengthening national and international security postures against unforeseen AI-related challenges.

Key insights

• Traditional government AI preparations rely on a 'predict-then-act' model, ranking risks by likelihood and impact.
• AI's inherent unpredictability makes likelihood-based risk assessment ineffective, as expert timelines vary significantly and official reviews confirm this limitation.
• A methodological framework is proposed for AI crisis preparedness based on capability-based planning, adapted from defense and homeland security.
• The framework includes three components: a systematically sampled scenario library, a capability rating procedure against these scenarios using coarse, gated criteria, and a prioritization method (partially described in the abstract).

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.18357