The National Institutes of Health canceled at least $2.6 million in grants tied to a single phrase — "structural racism" — according to court documents filed earlier this month in a lawsuit brought by University of California researchers. The filings from NIH, the National Science Foundation, the U.S. Department of Defense and the National Endowment for the Humanities reveal how federal agencies identified billions of dollars in research awards to terminate after President Donald Trump retook office: keyword searches, applied in bulk, without individual assessment.

The documents, first reported July 27, 2026, confirm that agencies cut grants if projects "expressed, or were presumed to express, viewpoints disfavored by the Administration."

What did the keyword lists contain?

NSF used the longest list of the four agencies. It told the court it pulled its search terms directly from a 2024 report by Texas Sen. Ted Cruz, which alleged the Biden administration had politicized science at the agency. The report contained hundreds of screening terms, including:

• "injustice"
• "ally"
• "prejudice"
• "institutional"
• "discrimination"
• "historically"
• "minority"
• "traumatic"

NSF also mentioned, but did not list, terms related to misinformation.

NIH worked from a shorter but similar list that included "workforce diversity" and "health equity," and said it also searched "various permutations" of each keyword.

The Defense Department searched for terms such as "pay equity," "LGBTQ" and "neurodiversity," and extended its net to energy and climate work. Grants referencing solar, wind or geothermal energy were targeted for termination, as were awards using terms like "climate change," "carbon neutrality," "decarbonization," "pollution control," "air quality management" or "federal sustainability."

NEH revoked grants based on keywords including "environmental justice," "transgenderism" and "diversity, equity, and inclusion" and its variants.

How systematic was the screening?

All four agencies told the court they identified awards to cancel "using general criteria" rather than assessing each grant individually.

The quality of disclosure varied. NSF provided a list of UC-held grants cut by this method but did not tie specific search terms to individual terminations. NEH likewise did not specify which keyword triggered each cancellation, though it sometimes shared portions of the flagged grant. NIH, in contrast, linked specific terms to cancellations in some cases.

One canceled NIH project at UC San Francisco studied how racism affected the mental health of adolescents and young adults who had contact with the legal system, and racism's impact on substance abuse. NIH flagged it for the phrase "structural racism." A second project at UCLA, focused on gestational diabetes among Asian Americans, was terminated for "health equality."

Dozens of additional NIH projects lost funding not because of specific phrases but because they carried diversity supplements — additional NIH-awarded funding intended to support researchers from underrepresented and disadvantaged backgrounds.

What is the legal stakes for R&D portfolios?

The UC researchers argue the terminations violate their First Amendment rights by targeting work based on viewpoint and subject matter. They have asked the federal judge overseeing the case to rule the revocations unconstitutional.

They have already won two preliminary injunctions blocking grant terminations made via form letters across several federal agencies. NIH told the court it specifically targeted grants held by researchers affiliated with UC's 10 campuses; in one example, the agency suspended over $500 million in funding to UCLA — funding a court has since ordered restored.

For research managers, the filings quantify the screening mechanics behind terminations that arrived as form letters with little explanation. The court's eventual ruling on constitutionality will determine whether keyword-based revocations can survive as an instrument of federal grant administration.