The Office of Management and Budget published a 412-page proposed rule in late May that would restructure how the federal government awards more than a trillion dollars a year in grants. Two provisions do most of the work.
Section 200.205 would require senior political appointees to review and approve discretionary grant awards before issuance, a checkpoint that did not exist in either prior era of federal grantmaking. Section 200.340 would expand authority to terminate any active award at any time for convenience, without cause or notice. The rule also restricts international scientific collaboration and adds preapproval requirements for conference and publication costs.
It applies government-wide, across HHS, NIH, NSF, HUD, DOJ, and FEMA among others. The comment period closed July 13, and OMB has targeted October 1 for the effective date.
The legitimate principle, stated fairly
There is a real argument on the government's side, and dismissing it makes the analysis worse.
Federal research money is public money, appropriated by Congress and administered by an executive branch accountable to voters. The proposition that career scientists and outside reviewers should hold effectively unreviewable authority over how a trillion dollars is spent is not self-evidently correct. Every democracy resolves the tension between technical expertise and political accountability somehow, and "the experts decide, full stop" is one answer among several. OMB's stated rationale is that the prior administration used grants to advance a policy agenda, which is a claim about political direction already existing rather than being newly introduced.
The useful question is therefore not whether political accountability belongs in grantmaking. It is where in the process it belongs.
The postwar American model located it at the top: politically appointed agency leadership, confirmed by the Senate, set institute budgets, research priorities, and program areas, while technical peer review determined which specific projects best advanced those priorities. Elected officials decided that cancer research deserved more money than it got last year. Scientists decided which cancer studies were methodologically sound.
This rule moves the political layer down to the individual award. That is the change, and describing it precisely matters more than characterizing it.
The effect that cannot be measured
Here is the part the debate mostly misses.
A review checkpoint at the end of a process changes behavior at the beginning of it. Researchers who know a political appointee will evaluate their application will adjust before submitting: avoiding certain topics, reframing questions, removing terminology, choosing a different study population. That adaptation happens across every application, not only the ones that would have been rejected.
The screening already underway makes the mechanism concrete. Nature reported that hundreds of NIH applications are held up at any given time by post-peer-review scrutiny, with some flagged by an algorithm for using terms such as "gender" and "climate change" among 235 disfavored terms.
Once a list like that is known to exist, researchers stop using the words. Then they stop framing questions in ways that would require the words. Then, eventually, they stop asking the questions. Each step is individually rational and none of it appears in any dataset.
That is the measurement problem. Denied grants can be counted. Terminated awards can be counted. Applications never written cannot be, and if the primary effect of a policy is on the composition of what gets proposed, the policy's main consequence is permanently unquantifiable. Anyone who later argues the rule had modest effects because rejection rates stayed low will be measuring the wrong thing.
Why the termination clause may matter more than the veto
The pre-award review has drawn the most attention. The provision allowing termination of any active award at any time for convenience may do more damage, because it operates on time horizons.
Research is a multi-year commitment. A five-year cohort study, a longitudinal trial, a training program producing scientists over a decade: these only make sense if funding is reasonably durable. Investigators hire staff, recruit patients, and commit graduate students on the assumption that a funded grant will be honored.
If any award can be canceled at any moment without cause, the expected duration of every grant becomes uncertain regardless of its nominal term. Rational actors respond by shortening their horizons. Shorter projects, faster results, less irreversible commitment, fewer people hired onto soft money, fewer dissertations built on federal awards.
The irony is specific. Long-horizon, high-risk research is precisely what federal funding exists to support, because private capital will not fund work with a fifteen-year payoff and no proprietary claim. A provision that selects against duration selects against the entire justification for public research funding, whatever anyone's political priorities. The same dynamic is visible in real time in a different corner of the federal health budget, where an appropriated CDC prevention program had its renewal declined by OMB months after Congress had already funded it.
Two conservative filters in series
There is a further compounding effect worth naming, because it is not obvious.
Peer review already has a well-documented conservatism bias. Reviewers favor proposals with preliminary data, established investigators, and predictable outcomes, which is why funders periodically create special mechanisms to support high-risk work that ordinary review would reject.
Adding a political review layer adds a second filter, and it also favors safety. A political appointee reviewing hundreds of grants faces asymmetric risk: approving something later characterized as objectionable carries career consequences, while blocking something valuable carries none, because the counterfactual is invisible.
Two risk-averse filters in series produce a substantially more conservative portfolio than either alone. Not because anyone intends it, but because each independently screens out the unusual.
Undefined standards maximize the effect
The rule reportedly instructs reviewers to ensure grants advance the President's policy priorities and do not promote "anti-American values," which is undefined.
Undefined standards in a review process are not merely a fairness problem. They are the configuration that produces the widest behavioral effect, because uncertainty about what is prohibited leads applicants to avoid anything that might be. A precise rule, even a restrictive one, lets people work around a known boundary. A vague one makes the safe strategy avoidance of the entire neighborhood.
If the goal is to prevent a specific category of spending, precision serves that goal better than ambiguity. Ambiguity serves a different goal.
What to watch
The rule is proposed, not final, and it may be narrowed, delayed, or enjoined.
The 45-day comment period for a 412-page rule affecting more than a trillion dollars is unusually compressed, and that will matter in litigation. Courts reviewing agency action under the Administrative Procedure Act examine whether an agency considered significant comments, weighed real-world impacts, and reasonably explained its choices. A short comment window on a long rule with substantial documented objections, including from the New England Journal of Medicine and numerous scientific societies, builds a record that cuts against the agency.
There is also an internal tension worth tracking. The proposal calls for grantees to conduct "Gold Standard Science," a term the administration has described only vaguely but which, in seeming contradiction to the OMB proposal, includes peer review. A rule that demotes peer review while invoking it as a standard has a coherence problem that opponents will press.
The measurable indicators, if the rule takes effect, are the ones to insist on: time from council recommendation to award, the share of awards delayed beyond historical norms, and the distribution of funded topics compared to prior years. None captures the self-censorship effect, but drift in the topic distribution is the closest available proxy. The same governance question sits underneath a separate story about federal research infrastructure, where a new nonprofit is trying to industrialize gene therapy development on ARPA-H funding, a model that depends on exactly the kind of durable, multi-year federal commitment this rule would put in question.
The deeper point holds regardless of where anyone sits politically. A system that funds research is also a system that signals what questions are safe to ask. Change the signal and you change the questions, long before you change a single funding decision, and the change will not appear in the numbers anyone thought to collect.