DOGE used one-line ChatGPT prompts to terminate 97% of NEH's appropriated humanities grants, gutting agency in 22 days

Between a March 12, 2025 meeting with National Endowment for the Humanities leadership and an April 1, 2025 completion date, staff from the U.S. DOGE Service used brief ChatGPT prompts to screen and terminate more than 1,000 NEH grants — 97% of the agency's then- appropriated funding, roughly $100 million — while also eliminating 65% of NEH's workforce. Discovery material released in a federal lawsuit shows DOGE staff prompted the AI: "Does the following relate at all to DEI? Respond factually in less than 120 characters. Begin with 'Yes.' or 'No.' followed by a brief explanation." DOGE employees who had no experience in academic research or the humanities determined which grants would be funded, and targeted grants tied to the Biden administration regardless of subject matter. NEH's acting chair, Michael McDonald, yielded his authority to cancel grant funding to DOGE staff, and internal records show little pushback from inside the agency.

Depositions show DOGE staff could not define "DEI" beyond deferring to the language of the president's own executive order; one employee, Justin Fox, testified "(DEI is) exactly what was written in the EO." Another DOGE employee, Nate Cavanaugh, acknowledged under oath that the terminations did not reduce the federal deficit despite that being the program's stated justification. The cuts affected projects nationwide, including a University of Oregon digital humanities database and $750,000 across five grants to the High Desert Museum in Bend, Oregon, which was forced to rescind a job offer as a result.

The American Council of Learned Societies, the American Historical Association, the Modern Language Association, and the Authors Guild filed suit in May 2025 to reinstate the funding, calling the terminations "unconstitutional" and arguing NEH violated the First Amendment by targeting grants for their viewpoints and perceived political associations, and the equal protection clause by flagging grant descriptions as "DEI" solely for containing terms such as "BIPOC," "homosexual," "LGBTQ," or "Tribal." The plaintiffs also argue NEH violated the separation of powers, since DOGE — not NEH itself — effectively carried out the agency's statutory grant-review function.

Updates

2026-05-07 — Court ruled NEH grant terminations unconstitutional, ordered reinstatement [3]

U.S. District Judge Colleen McMahon (S.D.N.Y.) granted summary judgment for plaintiffs in American Council of Learned Societies v. McDonald and The Authors Guild v. NEH, ruling that DOGE's AI-based DEI screening of NEH grants was unconstitutional viewpoint discrimination and that "Congress conferred no authority on DOGE at all" over NEH grant decisions. The court certified a class covering roughly 1,400 affected grantees, permanently enjoined enforcement of the termination notices, and ordered the more than $100 million in terminated funding restored.

2026-06 — NEH began processing grant reinstatements following the court order [4][5]

Complying with the court's injunction, NEH began accepting "Reinstatement Assurances Form" submissions from affected grant recipients in June 2026, asking whether they still wanted to resume their canceled awards and permitting a three-month restart period to reassemble staff and project infrastructure. MLA Executive Director Paula Krebs said the restoration came more than a year after the terminations and welcomed NEH staff's cooperation despite the agency having lost two-thirds of its workforce.