In an accompanying report, we found that two methods are best for verifying the location of AI chips: ping-based location verification (PLV) and physical inspections. This piece compares these two methods through a cost-benefit analysis and shows that PLV is the more cost-effective approach. We conclude that the most effective location verification approach would likely involve a PLV system that is supplemented by small numbers of physical inspections.

Large-scale AI chip smuggling undermines U.S. export controls that are designed to prevent adversaries from developing advanced AI capabilities. Location verification—which would help regulators determine where AI chips are located—has emerged as a key proposal to strengthen export control enforcement efforts. This report assesses the viability of location verification methods, their cost-effectiveness, requirements for implementation, and their value as an enforcement tool.

Immigrants make up nearly half of U.S. recipients of the world's most prestigious STEM awards, yet little is known about how this talent first arrives in the country. This data brief analyzes 550 recipients of the Nobel Prize, the Lasker Award, the ACM A.M. Turing Award, and MIT Technology Review's Innovators Under 35 between 2014 and 2024. Most foreign-born U.S. awardees first arrived as students, postdoctoral researchers, or visiting scholars—pathways now exposed to visa suspensions, entry restrictions, and proposed changes to optional practical training. Safeguarding these pathways is critical to maintaining U.S. leadership in STEM and emerging technologies.

Winning the AI race does not guarantee any economic, social, or military benefit. Without a theory of victory mapping AI development and deployment to competitive advantage, and ultimately to beneficial outcomes like profit or military victory, AI might provide more impressive capability than tangible benefit. This report evaluates various types of advantages that winning the AI race might provide and describes six theories of victory for converting those advantages into worthwhile outcomes.

How can we connect publication and patent data in new ways to derive useful insights into science and technology activity? How can we characterize these connections, visually or in natural language? This roundtable report summarizes expert responses to these questions and discussion of new approaches to analyze, connect, and visualize scientific publication and patent data to deliver insights and inform S&T decision-making.

CSET’s Jessica Ji shared her expert insight in an article published by Vox. The article examines growing concerns about catastrophic AI risks and the challenges of creating effective government oversight as AI systems become more capable and potentially harder to control.

CSET’s Helen Toner shared her expert insight in an interview with The New York Times. The conversation examines recent incidents involving AI systems hacking, deceiving humans, and coordinating with other AI agents, as well as concerns about the ability of companies to safely monitor and control increasingly capable AI models.

Despite decades of reform efforts, the federal government’s Authorization to Operate (ATO) process is a major barrier to delivering secure systems to warfighters at speed. This report examines why past ATO reforms have fallen short of their intent, and offers recommendations to getting leading American technology into the hands of warfighters faster and securely.

CSET’s Helen Toner shared her expert insight in an article published by The Washington Post. The article looks at recent incidents in which AI models from OpenAI, Anthropic, and Meta broke out of controlled testing environments and attempted to hack real systems, raising concerns about whether AI companies can safely control increasingly capable models.

AI systems are increasingly impressive, which makes their failures all the more baffling. This blog dives into the causes of AI misbehavior.