RAND Center on AI, Security, and Technology Quarterly Newsletter

We are pleased to announce that the Technology and Security Policy Center (TASP) and the Meselson Center are joining forces to become the RAND Center on AI, Security, and Technology (RAND CAST). This brings together RAND's cutting-edge work in two areas. TASP research focused on how high-consequence, dual-use technologies like artificial intelligence and biotechnology change the global competition and threat environment. Meselson concentrated on reducing risks from biological threats and emerging technologies. Together, the new RAND CAST will unite world-class policy analysis with leading technical expertise, giving decision-makers the insight they need to navigate global emerging threats.

As this merger took effect October 1, this newsletter presents work conducted under both centers. Going forward, you'll be able to find reports and commentary from both centers at rand.org/cast

As artificial intelligence advances, concerns have mounted that the technology could someday shape international politics in ways that raise the risk of conflict. In Six Ways AI Could Cause the Next Big War, and Why It Probably Won't, published in the Bulletin of the Atomic Scientists, RAND researchers Zachary Burdette, Karl P. Mueller, Jim Mitre, and Lily Hoak examine six pathways through which AI might contribute to the onset of future wars:

1. Balance-of-power shifts: If AI gives states significant military or economic advantages, it could convince leaders they had new opportunities for aggression.
2. Preventive wars: If leaders worry about a rival gaining large military or economic advantages over them, they might contemplate attacks to prevent those unwelcome shifts in the balance of power.
3. Lower costs of aggression: If AI significantly reduces human and material costs of war, AI could make the barriers to aggression appear lower.
4. Societal destabilization: If AI produces economic and social upheaval, leaders might look to deflect frustrations through external aggression.
5. Accidental actions: If AI takes unintended military actions through accidents or loss-of-control, it could inadvertently start or escalate crises.
6. Decision-making distortions: If AI decision-support tools give leaders incorrect or very hawkish advice, that could increase the risk of miscalculations and escalation.

Despite these concerns, the researchers conclude that such scenarios are unlikely to directly trigger a major war. The same dynamics that have historically restrained conflict remain in place: leaders tend to manage rather than escalate crises, nuclear deterrence discourages aggression, and safeguards such as institutions, human oversight, and crisis-management mechanisms further limit risk.

Read more: Six Ways AI Could Cause the Next Big War, and Why It Probably Won't

U.S. data centers today face a serious challenge: finding enough reliable electricity to power next-generation AI models. In their new report, Expanding U.S. Net Available Power Capacity by 2030, RAND researchers Ismael Arciniegas Rueda, Aisha Najera, Austin Smidt, Robin Wang, Hye Min Park, Henri van Soest, and David Gill explore the constraints to scaling the U.S. power grid to meet this overwhelming demand by 2030.

Examining literature and hosting discussions with experts, the authors identified 66 barriers to expanding grid capacity, grouping them into four categories: 1) permitting delays, 2) inefficient and costly interconnection processes, 3) underutilized transmission, and 4) a lack of supplemental power generation options.

The research team proposed actionable solutions for regulators, grid planners, and system operators — including streamlined federal permitting, enhanced transparency, incentives for adopting grid-enhancing technologies, and encouragement for supplemental non-diesel backup generation. If implemented, these recommendations could unlock 92 to 297 gigawatts of additional power capacity by 2030 — enough to support the coming wave of AI-driven demand.

Read more: Expanding U.S. Net Available Power Capacity by 2030: Barriers and Solutions

Artificial intelligence (AI) is transforming the life sciences, accelerating breakthroughs in research, drug discovery, and biotechnology. However, some of the AI tools that drive innovation can also be misused, posing significant dual-use risks. Ensuring these technologies are developed and deployed responsibly requires a clear, structured understanding of the capabilities of individual AI-enabled biological tools and their potential misuse applications.

RAND, in collaboration with the Centre for Long-Term Resilience, developed the world's first Global Risk Index for AI-Enabled Biological Tools. This flexible and comprehensive framework allows developers, funders, and policymakers to assess tools based on their capabilities, potential for misuse, accessibility, and technological maturity, supporting responsible development and deployment.

RAND researchers Toby Webster, Barbara Del Castello and Sana Zakaria, working with Centre for Long-Term Resilience collaborators Richard Moulange, James Walker, and Cassidy Nelson, evaluated more than 1,100 publicly available tools from 76 countries, ultimately shortlisting 57 examples that represented the frontier of capabilities for detailed analysis. Using a color-coding system to indicate levels of concern, they classified 13 tools as "red" — high priority for risk mitigation — and 15 as "amber", requiring case-by-case review. Many of the highest risk tools are fully open source; while this promotes innovation and access for legitimate researchers, it makes them equally accessible to threat actors.

The researchers include recommendations for how stakeholders can use the Index, such as assessing tools for misuse risks before development and deployment, funding the development of proportionate technical safeguards for data and models, and updating the Index regularly to maintain situational awareness among rapid advances.

Read more: Global Risk Index for AI-Enabled Biological Tools

The COVID-19 pandemic raised urgent questions about how the United States would fare under future biological threats, especially when it must rely on physical protective measures in the absence of early vaccines or other pharmaceuticals. In the new RAND report, Physical Approaches to Civilian Defense, a RAND research team examined three catastrophic pathogen scenarios:

• A "fast" pathogen that rapidly spreads from human to human.
• A "silent" pathogen that achieves widespread transmission before symptoms appear.
• A "saturating" pathogen that replicates in and permeates the environment, threatening human, animal, and plant life.

In analyzing each scenario, researchers evaluated the requirements and costs of tools such as highly protective respirators, air cleaning systems, pathogen-agnostic detection networks, and safe zone structures. Their findings underscore the importance of proactively identifying countermeasures to protect the U.S. vital workforce — those who maintain critical functions like healthcare, transportation, water, and energy — to prevent cascading failures during a biological event.

RAND's recommendations emphasize short- and long-term actions. To protect the U.S. population from "fast" or "silent" pathogens, they encourage policymakers to expand stockpiles of respirators and air cleaning systems, while a pathogen-agnostic detection system would provide early warnings of a silent spread. For a potential "saturating" pathogen, the team recommends more research and development on safe zone structures that could enable long-term habitation.

Read more: Physical Approaches to Civilian Biodefense