WASHINGTON D.C. – A former staffer in the Trump administration has testified under oath that the word “LGBTQ+” was used as a direct criterion to flag federal humanities grants for cancellation, with artificial intelligence being used to accelerate the process. The testimony provides a detailed look into a systematic effort to defund projects related to diversity, equity, and inclusion.
The revelations come from a deposition by Nathan Cavanaugh, a former political appointee who worked with the Department of Government Efficiency (DOGE), a team tasked with identifying federal programs for elimination. The deposition is part of a lawsuit filed by major academic organizations, including the American Historical Association, challenging the termination of over $100 million in grants from the National Endowment for the Humanities (NEH).
A Search for Keywords, Aided by AI
According to his testimony, Cavanaugh and a colleague, Justin Fox, were assigned to review hundreds of NEH grants. Neither had a background in academia or the humanities; Cavanaugh came from the tech startup world and Fox from investment finance. Their primary tool was a spreadsheet of grant descriptions, which they scanned for keywords associated with the administration's anti-DEI executive orders.
Court filings reveal that Fox used search terms like “gay,” “BIPOC,” “indigenous,” and “equality” to generate lists of grants for review. He also fed hundreds of grant descriptions into ChatGPT with a simple prompt: “Does the following relate at all to DEI?” Fox testified he did not provide the AI with a definition of DEI, leaving the model to interpret the term on its own.
The AI flagged numerous projects, often for merely mentioning marginalized communities. This included academic work on Black civil rights, a biography of a Black jurist, and an anthology of fiction by Jewish writers about the Holocaust.
‘It Explicitly Says LGBTQ’
The targeting of LGBTQ+ scholarship was particularly blunt. When asked why a grant for a public discussion series titled “Examining experiences of LGBTQ military service” was flagged for cancellation, Cavanaugh’s explanation was direct: “Because it explicitly says LGBTQ.”
Another project, which explored the legacy of HIV/AIDS activism and prison abolition through the lens of feminist and queer scholarship, was also flagged. Cavanaugh testified that the references to “queer insights,” “gender and LGBTQ studies,” and DEI were the reasons for its inclusion on the termination list.
Fox confirmed this approach in his own deposition, stating that their review was conducted through the “lens” of an executive order to eliminate spending on DEI. He noted that grants referencing LGBTQ+ topics were considered suspect because, in his words, “LGBTQ is often associated with underrepresented minority groups.”
The Aftermath
The lawsuit alleges that this review process led to the termination of more than 1,400 active NEH grants, totaling over $100 million and representing approximately 97% of the agency’s active projects. Termination notices were reportedly sent from an unofficial email address without individualized explanations.
The testimony also highlighted a complete lack of scholarly consultation. Cavanaugh admitted that he and Fox did not consult with academics or utilize the NEH’s established peer-review system. When pressed on what books or academic sources informed his judgment, he conceded, “There were no books.”
For communities in the Netherlands and across Europe, this case serves as a significant example of how political ideology, amplified by technology, can be used to systematically erase or defund the histories of marginalized groups. The use of AI to automate the targeting of specific communities is a development being watched closely by cultural and academic institutions worldwide.
