What 4,200 data points reveal about US government interference in research
Published here:
Since January 2025, many US federal actions affecting research have looked like isolated administrative decisions. A new report by #DefendResearch and Research Consulting, drawing on 244 academic and news sources, shows that together they form a consistent pattern. Rob Johnson outlines what the report found and what it means for research systems beyond the US.
Over the past two years, anyone following US research policy will have seen a steady stream of alarming stories . A dataset disappears from a federal website. A survey question is quietly dropped. A grant is cancelled mid-project. A government scientist is told not to publish without clearance.[1]
These changes have not gone unnoticed. Journalists, researchers and organisations have documented them in news reports, journal articles and commentary. Some have gone further: in early 2025, people across the scholarly communications and research communities founded #DefendResearch , a grassroots initiative to resist government censorship of academic research and defend academic freedom and scientific integrity.[2]
Taken individually, each of these changes can look like a routine administrative decision, a shift in priorities, or an isolated exception. A single account, however vivid, is easy to dismiss as anecdotal, and that critique has often been used to wave away concerns about political interference in science.
Working with #DefendResearch, Research Consulting set out to test whether these accounts add up to something more. Our report, From anecdote to evidence: Mapping the impact of US government action on public health research , draws on 244 academic records and news articles published between January 2025 and June 2026, and analyses more than 4,200 coded extracts. The methodology was preregistered on the Open Science Framework , and the search strings, coding decisions and underlying datasets are openly available, so others can check, challenge or extend the work.[3][4]
A deliberate pattern, not a series of exceptions
The evidence points to nine mechanisms operating at three levels of the research system.
At the system level, the public record itself was changed. One study found that 49% of 232 federal health datasets had potentially substantive changes, and only 13% of these were recorded in public documentation. Between January 2025 and January 2026, 562 federal information collections were discontinued, and 338 data collections dropped gender identity questions. Agencies and advisory bodies that produce evidence were cut back or dismantled entirely.[5][6][7]
At the institutional level, decisions on funding, expert advice and publication came under closer political control. An executive order now requires political appointees to review research grants for consistency with the administration’s priorities, and a further proposal would give them the final say over National Institutes of Health (NIH) and National Science Foundation (NSF) awards, with scientific review advisory only. A federal prosecutor questioned journal editors about bias , and the Health Secretary threatened to bar government scientists from publishing in leading journals .[8][9][10][11]
At the individual level , government scientists needed clearance from political appointees to publish or speak, and Centers for Disease Control and Prevention (CDC) manuscripts were withdrawn to remove terms such as “gender” . At the Food and Drug Administration, those who disagreed with the government’s new vaccine framework were told to “ submit your resignation letters ”.[12][13][14]
Fig.1: Nine mechanisms of interference identified in the report. Bars show how often each mechanism is documented across 244 sources, not how often it occurred. Mechanisms that leave little trace, such as self-censorship, are likely to be under-represented. Source: From anecdote to evidence (2026), CC BY 4.0.
Some effects of political pressure can be counted, such as terminated grants and deleted webpages, but others cannot: studies never proposed, questions never asked, findings never submitted. A research system can narrow what it knows without formally banning a topic, if pursuing that topic becomes risky enough.
Self-censorship may be the most widespread response of all, but it leaves the faintest record, and our evidence base almost certainly understates it. The same is true of the consequences for public health. Many will take years to emerge, and some of the data needed to measure them has itself been cut.
Recovery is not resilience
These actions did not go unchallenged. Researchers, institutions, courts, states, philanthropists and civil-society groups responded, often quickly: suing to restore data and funding, archiving threatened datasets before they disappeared, and publicly committing to protect editorial independence. Our report groups these responses into three strategies:
Reversal seeks to undo specific acts. Courts ordered about 100 Department of Health and Human Services (HHS) datasets restored and blocked a 15% cap on NIH indirect costs . But reversal restores the asset, not control over it, and it is open only to those with the standing and resources to litigate.[6][15]
Replacement builds alternatives. Philanthropies pledged nearly $1 billion , and civil-society networks mobilised within months to archive threatened data. But replacement preserves part of a function, not its scale or durability, and backups can go offline.[16]
Refusal means declining to alter or withdraw research in the first place. Journals and professional bodies have committed not to retract articles on political request . But these commitments are mostly pledges, and few have yet been tested.[17]
Fig.2: Three strategies for defending research, with examples of each and its main limit. Dots show how often each strategy is documented across 244 sources. Source: From anecdote to evidence (2026), CC BY 4.0.
Each of these responses is significant, but none of them changes the underlying problem: federal data, funding and scientific institutions remain dependent on executive decision-making. Infrastructure built over decades has been weakened in months, and it remains exposed.
Why this matters beyond the US
It would be comfortable to see this as a purely American story, but its effects reach much further. Researchers around the world rely on US federal funding and data, and on journals that have come under pressure. And because the US Department of Health and Human Services funds around 30% of all US university research and development , cuts there are felt by international collaborators too. As others have argued, Europe must learn from what has happened .[18][19]
Every research system rests on assumptions: that public data will stay accessible, that funding will be reasonably stable, and that researchers can pursue difficult questions without undue personal or institutional risk. These conditions tend to be taken for granted until they are tested . The US experience shows how quickly they can be worn away.[20]
Building a more resilient research system
The report makes eight recommendations, each aimed at the people best placed to act on it. Several apply well beyond the US:
• Preserve the record. Libraries and archives, in the US and abroad, should coordinate the archiving of federal data, with persistent identifiers, long-term hosting and mirrors held outside US jurisdiction.
• Set journal policies on government requests. Publishers should set out clear policies on how they handle government requests to withdraw, retract or alter papers, and record any such requests openly.
• Create shared backing for refusal. Professional bodies, funders and university associations should create shared legal and financial support, so that an institution’s ability to refuse political demands does not depend on how much it can afford to lose.
• Measure what is hard to see. Funders and learned societies should repeat and extend measures of self-censorship, combining surveys of researchers with analysis of grant and publication text.
I believe we may well see similar pressures in the UK and Europe in the coming years. If we do, the first signs will look like anecdotes: a missing dataset, a cancelled grant, a quiet word about what not to publish. The lesson from the US is to treat those early signs as a warning and prepare a response now. By the time the anecdotes add up to evidence, much of the damage will already have been done.
📑The report, executive summary and datasets are openly available at doi.org/10.26303/68s9-wa74 .[3]
This article gives the views of the author, not the position of LSE Impact or the London School of Economics. You are agreeing with our comment policy when you leave a comment.[21]