Federal agencies that terminated more than a thousand research grants at the University of California have admitted, in signed stipulations filed in federal court, that they used keywords related to diversity, gender, vaccine hesitancy and COVID-19 to screen for projects that ran afoul of the administration's priorities, CalMatters reported on July 21, 2026.
The admissions are in Thakur v. Trump, a proposed class action by UC researchers, filed in spring 2025 and pending before U.S. District Judge Rita Lin in San Francisco. According to the plaintiffs' July 15 motion for summary judgment, stipulations executed by NSF, NEH, the Defense Department, the Transportation Department and NIH-HHS state that each agency used “search terms, keywords, or phrases” and in some cases “artificial intelligence tools or software” to “identify grants for termination review,” and that each agency stipulated “the search terms, keywords, or other criteria used to determine which grants to terminate were designed to identify grants that expressed, or were presumed to express, viewpoints the Administration no longer wished to subsidize.”
CalMatters' review of the filings found that NIH used an internal search tool to flag UC grants mentioning “health equity,” “work force diversity,” “structural racism” and “sexual orientation,” and that the Transportation Department picked six grants worth about $42 million for termination because they sought “transportation equity” or prioritized “disadvantaged communities.” NIH alone suspended or canceled more than 1,000 UC grants.
The plaintiffs say the cancellations, which they put at nearly $2 billion, violated the First Amendment because the agencies targeted disfavored viewpoints rather than reviewing grants on their merits. They also argue that the Energy Department's termination of 283 grants in states that voted for Kamala Harris, while leaving similar grants in other states untouched, punished researchers for where they live. They asked the judge to rule without a trial, with a hearing set for October 20. A keyword screen does not judge whether a study is good science, only whether its text contains a disfavored term (see DOGE Gutted the National Endowment for the Humanities in 22 Days Using ChatGPT, Terminating 97% of Grants Including Holocaust Research and Massive Cuts to NIH and Cancer Research Funding).
Update (October 2, 2026): On August 18, the government asked the Supreme Court to review the Ninth Circuit's decision affirming the preliminary injunction for grants the plaintiffs say were cut in violation of the First Amendment. On September 1, Judge Lin approved a stipulation pausing summary-judgment briefing on every claim except the equal-protection claim over certain Energy Department grants while the Supreme Court petition is pending. Under the same agreement, NSF, NEH, the Defense Department and the Transportation Department agreed not to re-terminate the grants listed in their spreadsheets in the meantime.
Correction (October 2, 2026): The title said the administration admitted it killed “nearly $2 billion” in UC research “by keyword search.” The stipulations admit using keyword screens to select grants for termination review; the nearly $2 billion figure is the plaintiffs' total for the canceled grants. The title was revised.