Skip to content
FUTURECRAFT
← RAND Network

FC / NETWORK

Registry 01 · Domain

RAND Network · Research Domain

Artificial Intelligence

Seventy-five years of patient work, and a door held open for you.

Governance, safety, and the long-horizon institutional design of adaptive systems.

Brief / 01

Context record

AI governance at RAND is treated as an institutional design problem first and a technical one second, the central question is what kinds of oversight regimes remain credible as systems become adaptive, opaque, and embedded in critical decision-making.

RAND has been working on autonomous and AI-enabled systems for two decades, long before the contemporary frontier-model era. That work spans technical assurance, acquisition policy, military integration, civilian-sector governance, and the long-horizon implications of automating cognitive labor. Jason Matheny's tenure as President has further sharpened the institution's focus on frontier AI as a national-security and civilizational variable.

RAND's Signature Themes in Artificial Intelligence

Continuous evaluation regimes

Treating AI systems as adaptive artifacts that cannot be certified once and trusted forever, borrowing from how RAND has long approached nuclear posture and autonomous systems.

Compute and chip governance

Tracking concentration in compute, semiconductor supply chains, and the chokepoints that determine which actors can train frontier systems.

AI in national security

The integration of AI into intelligence, command-and-control, and defense acquisition, including the failure modes specific to high-stakes automation.

Workforce and economic effects

What forms of cognitive labor are augmented, displaced, or revalued as general-purpose AI diffuses across the economy.

Representative Work

  • Securing AI Model Weights

    RAND · RR-A2849-1

  • The Operational Risks of AI in Large-Scale Biological Attacks

    RAND · RR-A2977-2

  • AI and the Evolution of Biological National Security Risks

    RAND · RR-A2960-1

Methodological Emphasis

Continuous evaluation, structured analytic techniques, and red-teaming applied to systems whose behavior drifts under deployment.

Methodological Lineage →

Affiliated Researchers