Staff Reports
The federal judge’s blistering opinion on the National Endowment for the Humanities grant purge makes two things painfully clear: the administration rushed a mass policy change, and someone let ChatGPT play policymaker. Judge Colleen McMahon’s summary‑judgment ruling found the terminations unlawful and unconstitutional — and she didn’t mince words about how generative AI was used in the process.
What the judge found and why it matters
Judge Colleen McMahon granted summary judgment to plaintiffs challenging the mass termination of more than 1,400 NEH grants, finding the action violated the First Amendment and the equal‑protection component of the Fifth Amendment. The court called the agency action ultra vires — beyond lawful authority — and described the decision‑making as arbitrary and opaque. The opinion also flagged the use of ChatGPT by staff at the Department of Government Efficiency Service (DOGE) to identify grants tied to alleged “DEI” themes, noting that AI outputs were relied on with almost no human scrutiny.
AI in government: a bad idea done worse
Here’s the core problem: agencies are experimenting with AI for big, rights‑impacting decisions without rules or real oversight. DOGE staff asked ChatGPT to flag grants as “DEI‑related” and to spit out short “rationales,” then used those one‑line answers to pull the plug on roughly $100 million in funding. The court pointed out that models like ChatGPT can hallucinate and will often supply answers that match what a user seems to want — meaning the AI might simply have learned to echo the reviewers’ goal instead of providing reasoned analysis.
The administration’s execution failed conservative goals
Conservatives who favor reining in wasteful or ideologically bent grantmaking should not cheer sloppy execution. If your policy is to cut certain programs, you still have to follow the law. Letting a chatbot play “find the viewpoint” and firing grants on the basis of brief spreadsheet blurbs was a policy face‑plant. The result: a court restores rights and forces the administration to defend both the policy and the process on appeal. That’s not smart governing — it’s an invitation to lose in court and look arbitrary to the public.
What comes next: appeals, rules, and accountability
The government will almost certainly appeal to the Second Circuit and seek a stay. Meanwhile, the ruling should be a wake‑up call. If conservatives want durable reform of grantmaking and federal programs, they need better lawyering, clearer statutes from Congress, and ironclad procedures for using AI. Judges are rightly policing illegal action; but the smarter play is to build reforms that survive litigation. Use technology to help officials work faster, not to dodge legal checks or outsource judgment to a chatbot that has no idea about the Constitution.
