The White House wants federal funding to follow scientists, not institutions, while putting artificial intelligence and national security at the center of research. The approach could accelerate discovery—but weaken safeguards now supplied by universities and national laboratories.
The Trump administration wants to redirect federal research funding from established universities toward individual scientists, agile research teams and artificial intelligence.
The plan, detailed in Science: A New Golden Age, would reshape how agencies distribute and evaluate roughly $200 billion in annual research and development spending. Agencies with at least $3 billion in R&D authority must submit implementation plans within 90 days, according to the White House.
Many proposed tools already exist: portable fellowships, rapid grants, prize competitions, ARPA programs and regional technology hubs. The change is to make them central to federal science policy.
That could reduce bureaucracy and give researchers more independence. It could also separate scientists from the institutions that provide laboratories, cybersecurity, export-control compliance and protection against foreign technology transfer.
Federal grants usually go to universities, which employ researchers, administer funds and provide facilities. The White House wants more awards that scientists can carry between qualifying organizations.
Wall Street Journal reporter Amrith Ramkumar described the strategy as a transfer of support toward researchers and AI that could weaken universities dependent on federal grants.
White House science adviser Michael Kratsios told the Journal that federal science had stagnated by repeatedly funding similar work through the same mechanisms and institutions.
The report cites the National Science Foundation’s Graduate Research Fellowship and the NIH Director’s Pioneer Award as models. The Pioneer Award provides selected researchers with as much as $700,000 annually for five years without requiring a detailed experimental plan.
Portable awards could give scientists greater freedom to choose their institutions, supervisors and research questions. They could also help younger researchers escape laboratories in which senior investigators control funding and credit.
But research institutions supply more than administration. They maintain specialized equipment, animal-care programs, secure networks, conflict-of-interest reviews and systems for protecting human subjects. A portable grant cannot make a clean room, clinical-research operation or particle accelerator portable.
Moving money without replacing that support could leave researchers independent on paper but dependent on infrastructure the award no longer fully finances.
Expert panels now judge most federal grant applications. The White House would supplement them with rapid grants, prizes, advance market commitments and regranting programs. A “golden ticket” could let one reviewer advance an unconventional proposal rejected by the broader panel.
Stephen Ezell and Meghan Ostertag of the Information Technology and Innovation Foundation have cited evidence that conventional review can penalize novel and interdisciplinary work. In comments submitted to OSTP, they recommended longer investigator awards, high-novelty programs and partial lotteries when reviewers cannot distinguish among qualified proposals.
They also warned against weakening university technology transfer and argued that institutions often receive less than the full cost of federal research. That conflicts with the report’s concern that indirect-cost payments support administrative bloat.
The plan would require agencies to evaluate their complete funding portfolios, not only individual proposals. Internal “metascience” units would test grant mechanisms, track results and identify neglected fields.
NSF had already proposed such a unit for fiscal 2027. The new strategy would spread the model across government.
Loren DeJonge Schulman and Leya Mohsin of the Federation of American Scientists support federal metascience but caution that evaluation offices need authority, protected budgets and durable career paths. In a June 2026 analysis, they warned that evidence has little value if it arrives after decisions or officials lack incentives to use it.
A powerless metascience office could become another reporting requirement. An unaccountable one could give political appointees greater control over research.
That risk increases as the administration shifts from investigator-selected questions toward national missions.
Federal civilian science has traditionally preserved space for researchers to choose their subjects. The new strategy emphasizes explicit goals: AI-enabled discovery, quantum computing, fusion power, nuclear reactors, lunar exploration and advanced semiconductors.
This resembles the DARPA model. Program managers define problems, recruit teams, make several bets and terminate projects that miss milestones.
Mission-driven research is not new. NASA, the Energy Department and the military have long connected research to engineering and procurement. Congress also created ARPA-E, ARPA-H and NSF’s technology directorate.
The NSF Regional Innovation Engines already unite universities, businesses and governments around regional industries. Teams can receive as much as $160 million over a decade for research, commercialization and workforce development.
Earlier initiatives added mission programs to the traditional system. The Trump plan would make national competition—especially with China—an organizing principle for the system itself.
The report rejects the postwar idea that government funds basic research, universities make discoveries and industry later develops products. It argues that progress now moves repeatedly among science, engineering and manufacturing.
Scholars have challenged the old model for years. Venkatesh Narayanamurti, Tolu Odumosu and Lee Vinsel argued in Issues in Science and Technology that invention often enables discovery, not only the reverse.
MIT science-policy specialist William Bonvillian has described the United States as operating two systems: an integrated defense model extending through testing and procurement, and a fragmented civilian model focused on early research. In a 2022 essay, he argued that competition with China was already pushing the country toward industrial policy.
The White House plan combines existing bipartisan initiatives with greater skepticism toward university administration, stronger presidential direction and a central role for AI.
Current policy often treats AI as a research field or supporting tool. The new strategy treats it as scientific infrastructure.
As Ramkumar reported, the administration wants agencies to fund AI as an instrument of discovery rather than merely a way to improve existing tasks.
The Genesis Mission would connect national laboratories, supercomputers, scientific instruments and federal data with private AI and semiconductor companies. Autonomous laboratories could let software design experiments, operate equipment and interpret results.
The same systems could multiply errors. AI models can fabricate information and reproduce weaknesses in their training material. The report therefore calls for better data sharing, machine-auditable replication and AI-assisted verification.
Its credibility will depend on whether verification receives as much support as generation.
AI also creates a national-security problem. Scientific data placed in a commercial model, cloud system or shared repository may become available to foreign employees, contractors or collaborators. Remote access to laboratory equipment can expose operating knowledge even when the research results remain public.
These risks are greatest in the fields the plan prioritizes: semiconductors, quantum computing, nuclear technology, aerospace, biotechnology and advanced manufacturing.
Funding individual researchers does not change export-control law.
Under the Export Administration Regulations, releasing controlled technology or source code to a foreign person in the United States can be treated as an export to that person’s country of citizenship or permanent residence. The rule follows the technology and the recipient of the information—not the federal grant.
Most open academic research qualifies for the fundamental-research exclusion. Results ordinarily published and broadly shared, without proprietary or national-security restrictions, generally are not subject to the EAR.
That protection also follows the character of the research, not the institution conducting it. An independent scientist doing genuinely open work may receive the same treatment as a university researcher.
But the exclusion does not necessarily cover controlled source code, proprietary industrial data, equipment manuals or restricted technology supplied for the project. Publication reviews, citizenship restrictions or limits on dissemination can also move a project outside fundamental research. Defense-related technical data and services require a separate analysis under the International Traffic in Arms Regulations.
Universities and national laboratories employ export-control specialists who classify technology, screen collaborators, seek licenses and limit access to laboratories and files. National laboratories also operate stronger systems for classified information, controlled unclassified information and foreign-national access.
A scientist working through a startup, independent institute or temporary team may lack that support. The violation can occur before anyone recognizes that a license was needed.
Export control is only part of the problem.
Federal research-security policy relies heavily on institutions to verify foreign affiliations, disclose foreign support, train researchers and investigate conflicts. Under NSPM-33, organizations receiving more than $50 million annually in federal science and engineering support must maintain research-security programs.
NSF also requires disclosures and certifications involving foreign support and malign foreign talent programs. Its research-security framework seeks to protect federally funded research while preserving legitimate international collaboration.
Direct awards could improve personal accountability because certifications would follow the scientist. But they could also move research below institutional thresholds and replace professional compliance with self-reporting.
Researchers who change employers create another gap. A university can transfer a grant more quickly than it can transfer secure networks, laboratory access rules or export licenses. A collaborator approved at one institution may not be approved at the next.
The policy’s national-security safeguards must therefore become as portable as its funding.
Each direct award should designate a host institution or approved compliance provider responsible for:
Export classification and fundamental-research determinations.
Screening of collaborators, visitors and foreign organizations.
Technology-control plans and access restrictions.
Review of cloud, AI and cross-border data systems.
Licensing, training and recordkeeping.
Reassessment when the researcher, institution or project changes.
Reporting suspected unauthorized releases.
Open fundamental research could follow a lighter process. Projects using proprietary, defense-related or otherwise controlled technology would require stronger review before an award or transfer.
Agencies could finance shared compliance services for independent researchers and small organizations. NSF’s SECURE Center offers one model for distributed research-security support, although export compliance would require direct involvement from the Commerce and State departments.
Awards involving controlled technology should not transfer automatically. Portability should pause until the new host demonstrates equivalent safeguards.
The White House plan addresses real weaknesses: slow grants, administrative burdens, weak support for unconventional work and gaps between discovery and manufacturing.
Its risks concern power and capacity. Who chooses national priorities? How much control should political appointees have over grants? Can money leave universities without damaging research infrastructure? Can individual awards preserve export control, cybersecurity and research security? Will AI accelerate reliable science or merely its output?
Congress still controls appropriations, and many changes require legislation. But the administration has tied its strategy to agency plans and the fiscal 2028 budget.
Funding scientists directly need not weaken national security. It will do so, however, if the government makes money portable while leaving compliance behind.
The measure of the policy will be where the money moves, what it produces—and who gains access along the way.
Comments
No comments on this item Please log in to comment by clicking here