Court documents revealed that DOGE employees used ChatGPT to review over 1,000 National Endowment for the Humanities grant proposals, and the AI chatbot flagged a $349,000 grant to replace the aging HVAC system at High Point Museum in North Carolina as DEI. ChatGPT's reasoning: “Improving HVAC systems enhances preservation conditions for collections, aligning with the goal of providing greater access to diverse audiences.” The grant was cancelled.
DOGE employees Justin Fox and Nate Cavanaugh uploaded 1,163 NEH grant proposals into ChatGPT to identify DEI-related projects. Of those, 1,057 were flagged as DEI-related and cancelled. Only 42 were retained. Over $100 million in NEH funding was cut based on ChatGPT's determinations, representing approximately half the agency's yearly budget.
The revelation emerged through discovery in a lawsuit filed by the American Council of Learned Societies, the American Historical Association, the Modern Language Association, and the Authors Guild, arguing that DOGE violated First Amendment rights and equal protection guarantees by making funding decisions based on AI-flagged characteristics. High Point Museum Director Edith Brady confirmed the grant was terminated after the project had already begun, though the museum recovered approximately 70% through a termination clause.
The federal government effectively outsourced the decision of which humanities projects deserve funding to ChatGPT, an AI tool known to hallucinate facts and make errors of reasoning. It concluded that fixing a building's air conditioning constitutes a diversity, equity, and inclusion initiative. This is what happens when you replace expert judgment with a chatbot.