type: timeline_event
On March 19, 2026, court documents filed in the ongoing NEH grant cancellation lawsuit revealed one of the most absurd examples of DOGE's AI-driven decision-making: a $349,000 grant to the High Point Museum in North Carolina to replace its aging HVAC system had been flagged by ChatGPT as "DEI-related" and subsequently canceled. The AI's reasoning, preserved in the court record, was that improved climate conditions in the museum would support "greater access to diverse audiences" — a logic chain that transformed a basic infrastructure maintenance project into a prohibited diversity initiative.
The HVAC grant became an instant symbol of the broader dysfunction in DOGE's grant review process. Of the 1,163 active NEH grants evaluated by ChatGPT, 1,057 were flagged as DEI-related — a 97 percent cancellation rate that critics said demonstrated the tool was not performing meaningful analysis but rather rubber-stamping the elimination of nearly every grant in the portfolio. No definition of "DEI" had been provided to the AI, no parameters were established for what would constitute a legitimate versus problematic grant, and no human being reviewed the AI's determinations before cancellation notices were sent.
The High Point Museum, a small community institution in a city of 115,000 people, had been counting on the grant to replace HVAC equipment that was failing and threatening the preservation of its historical collections. Museum officials said the cancellation put irreplaceable artifacts at risk and that they had no alternative funding source for the project. The museum's plight was representative of hundreds of similar institutions across the country that had lost funding through the same automated process.
The revelation underscored what critics called the fundamental absurdity of using a consumer AI chatbot to make consequential federal funding decisions. ChatGPT, designed for general conversation and not for policy analysis, had no understanding of federal grant categories, no ability to distinguish between infrastructure maintenance and programmatic content, and no framework for evaluating whether a grant advanced diversity goals. Yet its outputs had been treated as authoritative determinations warranting the cancellation of hundreds of millions of dollars in federal support for the arts, humanities, and cultural preservation.