The testimony of a young political appointee has pulled back the curtain on a contentious review of federal humanities funding. In a January deposition, Nathan Cavanaugh described how members of the Department of Government Efficiency, known as DOGE, working through the General Services Administration reviewed hundreds of grant summaries from the National Endowment for the Humanities and identified projects that mentioned LGBTQ communities or related topics. Those brief project descriptions were used as primary signals in a process driven by executive directives targeting DEI priorities, with DEI here understood as diversity, equity, and inclusion. The result: many awards were terminated rapidly and with little public explanation.
The NEH normally funds a wide range of work—archives, museum programs, scholarly research and public conversations—through a peer review system that academic organizations rely on. But plaintiffs in a lawsuit including the American Council of Learned Societies, the American Historical Association and the Modern Language Association argue that this review was bypassed. Cavanaugh’s deposition, released publicly on video, offers granular examples of how a handful of reviewers scanned short descriptions and marked projects for cancellation, sometimes apparently because a single word appeared in a synopsis, and sometimes after running text through an AI tool.
What the deposition described
Cavanaugh and a colleague, Justin Fox, were assigned to a DOGE team that focused on small agencies and grant portfolios. Neither reviewer came from the humanities: Cavanaugh had built startups and worked in tech; Fox came from finance. Their task, under instructions to identify federal spending tied to DEI or other programs the administration considered wasteful, consisted largely of reading brief project descriptions in spreadsheets. According to the deposition, the reviewers looked for language such as “gay,” “BIPOC,” “indigenous,” and other markers that might signal a link to race, gender, or sexual orientation. The approach placed a premium on short summaries rather than academic reviews or subject-matter expertise.
How grants were identified and processed
One striking element of the process was the use of automated tools alongside manual scanning. Justin Fox reportedly ran hundreds of descriptions through ChatGPT, asking the model whether a description related to DEI, even though no precise definition was provided to the system. The AI frequently flagged projects that referenced marginalized groups or historical injustices. Examples cited in filings include a public discussion series on the experiences of LGBTQ military service members and a project on the legacy of HIV/AIDS activism tied to prison abolition. In some cases, the mere presence of words like LGBTQ or “gender” in a description was enough to trigger scrutiny and ultimately cancellation.
Use of AI in screening
The introduction of ChatGPT into grant review raises procedural questions: how did the model interpret context, and were results audited by human experts? Deposition testimony indicates that reviewers did not supply a formal definition of DEI to the AI and acknowledged limited understanding of how the tool reached its determinations. As a consequence, projects addressing matters such as the Colfax Massacre of 1873, biographies of Black jurists, and translated works by Jewish writers reflecting on the Holocaust were reportedly flagged simply because their descriptions engaged with race, religion, or marginalization in some form. Those flags formed the basis for termination notices sent outside the agency’s standard systems.
Who made the final calls and the fallout
Court documents allege that more than 1,400 active NEH grants—representing over $100 million—were terminated, a figure described as roughly 97 percent of the agency’s active awards at one point. Some termination letters were sent from unofficial email accounts rather than through the NEH’s normal grants portal, and communications show agency leaders warning that many cancellations lacked adequate justification. Acting NEH Chair Michael McDonald, for example, expressed concern internally even as the record indicates that DOGE retained decision authority. The plaintiffs say the cancellations were unlawful and deprived scholars and cultural organizations of due process and vital funding.
Implications for oversight, expertise, and legal review
The deposition also underscored a broader governance issue: decisions about scholarship were made by reviewers who did not consult NEH’s peer review panels or subject experts. Cavanaugh conceded that judgment calls were made without reference to academic books or formal humanities credentials, at one point admitting there were “no books” consulted to inform their choices. The legal challenge brought by scholarly associations contends that terminating grants on the basis of terse descriptors and algorithmic flags violates statutory and administrative safeguards. The White House declined to comment on the record, and the dispute now turns on judicial review of process as much as policy.
What remains at stake
The controversy raises immediate questions about how federal grantmaking will balance political directives, institutional expertise, and technological shortcuts going forward. For universities, cultural institutions and small organizations that rely on NEH funding, the case highlights the vulnerability of projects that engage with race, gender and sexual orientation. For policymakers and administrators, it prompts a debate about the role of AI, the meaning of DEI in public spending, and the safeguards needed to protect rigorous peer review and transparent decision-making in federal research funding.