How Federal Keyword Screening Killed Billions in Research Grants

After President Trump returned to office in 2025, four federal agencies built keyword-screening systems to hunt down and terminate research grants tied to disfavored topics — chiefly DEI and climate work. Court filings forced the National Science Foundation, National Institutes of Health, Department of Defense, and National Endowment for the Humanities to disclose the actual term lists. The NSF’s was the broadest, lifted from a 2024 report by Senator Ted Cruz alleging the Biden administration had politicized science, and flagged words as generic as ‘ally,’ ‘institutional,’ ‘historically,’ ‘minority,’ and ‘traumatic.’ The other agencies targeted terms like ‘health equity,’ ‘LGBTQ,’ ‘neurodiversity,’ ‘transgenderism,’ ‘environmental justice,’ and a cluster of climate phrases including ‘carbon neutrality’ and ‘air quality management.’

Because the screening keyed on isolated words rather than a project’s substance, it swept up research with no ideological content. A UCSF study on racism and adolescent mental health was cut for the phrase ‘structural racism’; a UCLA gestational diabetes project was killed over ‘health equality.’ NIH alone suspended more than $500 million tied to UCLA, with grants scrapped mid-project — a pattern that maps closely to keyword hits rather than any review of scientific merit.

The episode matters as a case study in crude automated filtering driving high-stakes decisions: a lexical match, not human judgment, determined which science lived or died. University of California researchers sued, arguing the form-letter terminations amounted to viewpoint-based discrimination in violation of the First Amendment, and won two preliminary injunctions that blocked the cancellations across several agencies and restored some funding.