US court criticizes the use of ChatGPT in public grant cuts
A federal judge found humanities grant cuts linked to DOGE unconstitutional. The case has become a strong example of a core rule: using AI does not remove human responsibility.
Summary
A US federal court found more than $100 million in cuts to National Endowment for the Humanities grants unconstitutional. According to the ruling, DOGE used ChatGPT to help identify projects allegedly linked to diversity, equity, and inclusion, without sufficiently clear criteria or adequate review.
Judge Colleen McMahon concluded that the cuts involved viewpoint discrimination and violated constitutional protections. Affected projects included work related to Holocaust education, Indigenous communities, civil rights, and African American history.
In practice
The central issue is not simply that an AI tool was used. It is that AI was used to classify projects with real consequences, without a robust definition of the criteria and without enough human responsibility for the outcome.
For companies and public bodies, the lesson is direct. AI systems can support decisions, accelerate analysis, and help organize information. But decisions with financial, legal, employment, or social impact need explicit criteria, human oversight, documentation, and appeal mechanisms.
Context
AI adoption in government and business is moving faster than many internal governance models. General-purpose tools such as ChatGPT are easy to use, but that also increases the risk of poorly framed decisions, especially when the output appears objective or technically neutral.
The DOGE case shows that responsibility does not disappear when an organization uses AI. If an automated or semi-automated decision harms people, projects, or rights, the organization that applied it remains responsible for the criteria, process, and consequences.
Why it matters
- It is a concrete example of legal risk in the use of AI for public decisions.
- It shows that “the AI did it” should not work as a shield against responsibility.
- It reinforces the need for internal policies on AI use in sensitive decisions.
- It is a useful case for enterprise training on governance, audit, and human oversight.
The practical conclusion is simple: AI can help with decision-making, but it should not replace the duty to justify. The greater the impact of a decision, the stronger the human process around the tool must be.