U.S. Dual-Use Research Oversight Imposes Domestic Costs While Gaps Persist Globally, Study Finds

A new empirical paper in Science is the most data-intensive examination yet of a problem biosecurity analysts have argued for years: that single-country oversight of dual-use research may penalize domestic scientists without actually closing the global supply of dangerous knowledge.

Published June 5, Professor Seokbeom Kwon of KAIST's School of Business and Technology Management analyzed roughly 600,000 research papers to map the relationship between security oversight and scientific output. According to the KAIST press release reviewed by EurekAlert, the study found that tightening oversight by one country alone carries disproportionate costs for domestic research while doing little to stop comparable work from advancing elsewhere. The paper, titled "Dual-use research under scrutiny" (DOI: 10.1126/science.aee2479), is the first large-scale empirical treatment of this tradeoff.

The dual-use research of concern (DURC) category covers work that has legitimate civilian applications, vaccine development, transmission studies, pathogen characterization, and potential misuse for bioweapons or bioterrorism. That classification is not contested here. What the KAIST study puts numbers behind is the effectiveness confidence around unilateral oversight as a control mechanism. The confidence on that effectiveness, per the findings, should be rated low.

The U.S. regulatory posture has been moving in a more restrictive direction. Executive Order 14292, signed in May 2025, intensified federal oversight of biological research with security implications, including gain-of-function work. But as Kwon's analysis flags, existing oversight frameworks like National Security Decision Directive 189 apply primarily when federal funding is involved, which means privately funded or foreign-conducted research falls outside their reach. That is not a new observation, but attaching a 600,000-paper dataset to it gives policymakers something concrete to argue from.

The timing matters because the institutional apparatus the U.S. would need to do better is itself under strain. According to a May 2026 analysis by CSIS's Wadhwani AI Center, reviewed in the published PDF on the CSIS website, the administration's FY 2026 budget proposal threatens cuts of $18 billion to NIH, $3.6 billion to the CDC, and $240 million to the Administration for Strategic Preparedness and Response. The same analysis notes the White House Office of Pandemic Preparedness and Response Policy is currently leaderless following the resignation of its top official, Gerald Parker.

Layered on top of those biosecurity-specific cuts is a parallel problem in the AI-biosecurity space. CAISI, the Center for AI Standards and Innovation within NIST, holds the government's primary mandate for evaluating AI systems that could pose biosecurity risks. NIST's FY 2026 budget faces a proposed cut of roughly $325 million, a nearly 30 percent reduction, according to the CSIS analysis. A separate assessment published by the Institute for Progress estimates that CAISI is currently operating on roughly $15 million annually and would need at least $84 million per year to execute all AI Action Plan taskings assigned to it. The biosecurity evaluation function, assessing whether AI models can provide meaningful uplift to someone trying to engineer a dangerous pathogen, sits inside that underfunded portfolio.

The dual-use dilemma has always been about ambiguity that can't be resolved at the point of the experiment. Research on viral transmission mechanisms or pathogen behavior has produced both vaccines and, in other hands, weapons. The KAIST finding doesn't argue against oversight; it argues that oversight architecture designed around federal funding triggers and U.S. institutions is structurally mismatched to a research ecosystem that is global, partly private, and increasingly AI-assisted.

Kwon's conclusion, as reported in the EurekAlert release, is that international cooperation and balanced policy design are necessary to reduce those structural tensions, a recommendation that the current trajectory of U.S. biodefense funding makes harder to act on, not easier. There's no analytic confidence level at which cutting the agencies responsible for screening, surveillance, and AI risk assessment while simultaneously tightening domestic research rules produces a net security gain.

Sources cited:
- EurekAlert (KAIST press release) (https://www.eurekalert.org/news-releases/1131017)
- Science, "Dual-use research under scrutiny," Kwon (2026) (https://www.science.org/doi/10.1126/science.aee2479)
- CSIS Wadhwani AI Center, "Opportunities to Strengthen U.S. Biosecurity from AI-Enabled Bioterrorism" (https://www.csis.org/analysis/opportunities-strengthen-us-biosecurity-ai-enabled-bioterrorism-what-policymakers-should)
- Institute for Progress, "What Will It Cost for the US to Be Ready for the Next Big AI Breakthrough?" (https://ifp.org/funding-for-caisi/)
- Homeland Security Newswire, dual-use research coverage (https://www.homelandsecuritynewswire.com/dr20260608-first-largescale-empirical-analysis-of-dualuse-research-and-security-oversight)