ChatGPT Behind $100M Humanities Funds Slashed

A lawsuit discovery reveals that DOGE employees used ChatGPT to flag and terminate over $100 million in humanities grants, cutting funding for Holocaust education, Native American language preservation, and Italian American archives while labeling projects focused on Jewish culture as “DEI-related.”

Story Snapshot

DOGE staffers used ChatGPT and keyword searches to eliminate 97% of NEH’s grant pipeline, totaling over $100 million
Grants supporting Holocaust survivor stories, Native American languages, and Appalachian history were terminated despite NEH leaders confirming they weren’t DEI-related
DOGE operated without legal authority, communicating via Signal app in violation of federal records requirements
Shortly after the cuts, a $10 million single-source award was given to the conservative Tikvah Fund

DOGE’s AI-Driven Grant Termination Process

Justin Fox, a DOGE Small Agencies Team member, led the effort to identify and eliminate what he termed “radical DEI” grants at the National Endowment for the Humanities. According to depositions released in March 2026, Fox created ChatGPT prompts that flagged grants containing keywords like “BIPOC,” “LGBTQ,” and “Tribal.” The AI system categorized projects focused on Jewish cultures and Holocaust education as DEI-related. Fox admitted in his deposition that he never formally defined what constituted a DEI grant, yet he made final termination decisions affecting over 1,000 grants worth more than $100 million. The process occurred between March and April 2025, eliminating 97% of NEH’s grant pipeline.

Questionable Targeting of Historical and Cultural Projects

The terminated grants supported diverse educational and preservation initiatives across American communities. Projects documenting Soviet Jewish women’s experiences during the Holocaust were cut despite NEH Acting Chair Michael McDonald testifying that Holocaust-related work should not be classified as DEI. Italian American archival projects, Appalachian photographic collections, and Native American language preservation programs also lost funding. Authors Guild CEO Mary Rasenberger characterized the process as “ridiculous,” describing DOGE staffers as inexperienced personnel “being told just go in and cut as much as you can.” Even infrastructure grants for HVAC upgrades were terminated. McDonald objected to numerous cuts but was overridden by DOGE operatives who lacked congressional authority to make such decisions.

Federal Records Violations and Procedural Concerns

Discovery documents revealed that DOGE staff conducted their grant review using the Signal messaging app, which does not preserve communications as required by the Federal Records Act. The team operated through non-government email accounts and communicated outside official channels, raising serious transparency and accountability questions. McDonald testified he was unaware ChatGPT was being used to evaluate grant applications. The lawsuit, filed by the American Council of Learned Societies, American Historical Association, Modern Language Association, Authors Guild, and affected scholars, alleges constitutional violations including viewpoint discrimination and violations of separation of powers. Plaintiffs note that DOGE’s keyword searches included terms associated with minority groups but excluded terms like “white” or “heterosexual,” suggesting biased criteria.

Aftermath and Ongoing Legal Challenge

The grant terminations resulted in 65% of NEH staff losing their positions and halted critical humanities projects supporting schools, libraries, and community organizations nationwide. Shortly after the cuts, McDonald solicited and awarded a $10 million single-source grant to the Tikvah Fund, a conservative Jewish organization, without the competitive review process typically required for NEH funding. The lawsuit seeks reinstatement of the terminated grants and compensation for affected institutions. As of March 2026, plaintiffs filed a motion for summary judgment to avoid trial, presenting depositions and internal communications as evidence of unlawful government overreach. The case remains pending judicial review, with implications that could affect how executive agencies exercise authority over congressionally appropriated funds and whether AI-driven decision-making can substitute for expert human judgment in federal grant administration.