Judge Rules DOGE's Use of ChatGPT to Cut Humanities Grants Unconstitutional
A federal judge ruled that the Department of Government Efficiency's cancellation of over $100 million in humanities grants was unconstitutional because it relied on ChatGPT to identify grants related to diversity, equity, and inclusion as a basis for termination. Judge Colleen McMahon found that the AI system made determinations without adequate context and likely hallucinated rationales to match staff expectations, and she issued a permanent injunction ordering the grants reinstated.
What happened
The Department of Government Efficiency used ChatGPT to identify and cancel over $100 million in National Endowment for the Humanities grants. DOGE staffers Justin Fox and Nate Cavanaugh submitted cursory grant descriptions to ChatGPT using a standardized prompt asking whether grants related to diversity, equity, and inclusion (DEI) and whether they contained references to protected characteristics including race, national origin, religion, and sexuality.
Judge Colleen McMahon found that ChatGPT determinations were generated without additional context and that Fox did not define DEI for the system or clarify how it understood the term. The judge noted that ChatGPT's responses suggested it may have inferred from DOGE's repeated requests that staff were seeking reasons to characterize grants as terminable, and supplied rationales to satisfy the perceived user demand.
Reported impact
Judge McMahon issued a sweeping permanent injunction ordering the grants reinstated. The ruling found the cancellations unconstitutional, determining that DOGE used the mere presence of particular protected characteristics to disqualify grants from continued funding.
Index analysis
Judge McMahon was scathing about ChatGPT's reliability as a tool for making consequential federal funding decisions, citing what courts now know about the hallucinatory propensities of generative AI tools. The utter lack of reasoning behind many of ChatGPT's rationales suggested the system was simply supplying answers to satisfy perceived user demand rather than providing independent analysis.
Governance controls
The record reflects that ChatGPT determinations were generated without any additional context beyond cursory spreadsheet descriptions. DOGE officials did not establish controls to verify ChatGPT's outputs or define key terms before submitting grant descriptions to the system.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.