Regulatory Consulting at RAND Corporation

Overview

RAND is a nonprofit, nonpartisan research institution with roughly 2,000 researchers and 75 years of national-security and domestic-policy work, and it manages several federally funded research and development centers (FFRDCs) for the U.S. government. Its AI-policy work is anchored in the RAND Center on AI, Security, and Technology (CAST), part of RAND Global & Emerging Risks, which unites world-class policy analysis with genuine technical expertise on frontier models, compute, and catastrophic risk. That combination — plus RAND’s tradition of client-sponsored analysis for the Department of Defense, the intelligence community, and state governments — makes it one of the few institutions that is simultaneously a think tank, a technical research shop, and a trusted government advisor. RAND’s institutional credibility on AI is amplified by its leadership: President and CEO Jason Matheny is one of the country’s foremost authorities on the intersection of technology and national security, having previously led White House technology and national-security policy at the NSC and OSTP, served as a commissioner on the National Security Commission on Artificial Intelligence, directed IARPA, and founded Georgetown’s CSET. Under his tenure, RAND has advanced a distinctive governance thesis — that oversight should focus on three chokepoints in the AI supply chain (hardware, model training, and rigorous pre-release review) — and CAST has become a leading voice on securing model weights, mitigating catastrophic risk, and the co-evolution of AI, cyber, and national security.

AI-Focused Regulatory Work

RAND’s AI-policy weight comes from a rare stack: methodological rigor forged ‘in the shadow of catastrophe,’ deep national-security credibility, and a technical center willing to get its hands dirty with the machinery of frontier AI. The clearest expression is CAST’s ‘Securing AI Model Weights,’ which laid out concrete recommendations for protecting the weights of frontier models from theft and misuse — work that has shaped how Washington thinks about the security of the most powerful systems. Founding CAST director Sella Nevo came to RAND from Google, where he’d built flood-forecasting systems covering hundreds of millions of people, and he now leads a program spanning AI security, technology-policy governance, and biosecurity, including Canary, a RAND–METR initiative advancing safer frontier AI through rigorous risk evaluation. What sets RAND apart from advocacy-driven shops is its posture of quality and objectivity: its 2026 Delphi study on legal and policy approaches to catastrophic AI harms notably found limited expert optimism about comprehensive near-term federal action, pointing instead toward state-level initiatives and voluntary measures — a sober conclusion RAND published rather than buried. Jason Matheny’s own congressional testimony has repeatedly pressed a supply-chain model of AI oversight and warned about AI-enabled cyber and disinformation at unprecedented scale. For someone building a career in AI governance, RAND is also a genuine on-ramp: its CAST Fellowship recruits from undergraduates through mid-career professionals across AI security, technology-policy governance, and biosecurity, based in DC, San Francisco, or Pittsburgh, or remote.

Notable AI Policy Work

• Published ‘Securing AI Model Weights: Preventing Theft and Misuse of Frontier Models’ (Nevo et al., RR-A2849-1) — a widely cited blueprint for protecting frontier-model weights that has influenced federal thinking on model security.
• Released ‘Legal and Policy Approaches to Mitigate Catastrophic Harms from AI’ (2026, RR-A4266-1), a Delphi study with 24 AI-policy experts finding limited near-term federal action likely and emphasizing state-level and voluntary measures.
• Published Richard Danzig’s ‘Artificial Intelligence, Cybersecurity, and National Security: The Fierce Urgency of Now’ (2025, PE-A4079-1), analyzing U.S. government failings at the AI–cyber intersection.
• Operates Canary, a RAND–METR initiative advancing safer frontier AI through rigorous risk evaluation and biosecurity research, and runs a CAST quarterly newsletter tracking the field.
• Delivered repeated congressional testimony (CEO Jason Matheny) advancing a three-chokepoint AI supply-chain oversight model — hardware, training, and pre-release review — before the House Science Committee, HPSCI, and Senate committees.
• Runs the RAND CAST Fellowship, a structured pipeline into AI-security, tech-policy-governance, and biosecurity research for candidates from undergraduate to mid-career level.

Thought Leadership

1. RAND Center on AI, Security, and Technology (CAST)
   The RAND Center on AI, Security, and Technology (CAST) — cutting-edge research on frontier-AI security, compute, and catastrophic risk, plus the Canary (RAND–METR) evaluation initiative.
2. Legal and Policy Approaches to Mitigate Catastrophic Harms from AI (2026)
   A 2026 Delphi study of 24 experts on legal and policy approaches to catastrophic AI harms — finding comprehensive near-term federal action unlikely and emphasizing state/voluntary measures.
3. AI, Cybersecurity, and National Security: The Fierce Urgency of Now (2025)
   Richard Danzig’s 2025 perspective on the urgent AI–cybersecurity–national-security nexus and U.S. government failings to prepare for it.
4. Advancing Trustworthy AI — Jason Matheny congressional testimony
   CEO Jason Matheny’s congressional testimony on advancing trustworthy AI and a three-chokepoint (hardware, training, pre-release review) supply-chain oversight model.
5. Challenges to U.S. National Security and Competitiveness Posed by AI (testimony)
   Matheny’s Senate testimony on the national-security and competitiveness challenges posed by AI.
6. RAND CAST Fellowship
   The RAND CAST Fellowship — a structured pipeline into AI-security, tech-policy-governance, and biosecurity research (undergraduate to mid-career).

Clients

RAND manages FFRDCs and conducts sponsored research for the Department of Defense, the U.S. intelligence community, the Department of Homeland Security, and state governments; CAST’s technology-and-security work is supported by philanthropic funders alongside government sponsors, with mandatory funding disclosure to preserve independence.