The Department of Government Efficiency promised to bring Silicon Valley's move-fast-and-break-things ethos to the federal bureaucracy. On Thursday, a federal judge made clear that some things—like constitutional due process—are not meant to be broken.
US District Judge Colleen McMahon issued a 143-page ruling declaring DOGE's cancellation of over $100 million in federal grants unconstitutional. The decision is notable not just for its outcome but for its withering documentation of how the cuts were made: officials fed grant descriptions into ChatGPT and used the chatbot's outputs to justify eliminations, with minimal human review of whether the AI's assessments bore any relationship to reality.
The process was the problem
Judge McMahon's ruling doesn't merely object to the use of AI in government decision-making—it objects to the specific manner in which DOGE deployed it. According to the decision, staffers prompted ChatGPT with grant summaries and asked it to flag programs that appeared duplicative, wasteful, or misaligned with administration priorities. The model's responses were then treated as sufficient justification for termination, even when grantees had existing contractual protections and statutory rights to their funding.
The judge found this process violated the Administrative Procedure Act's requirements for reasoned decision-making and the Fifth Amendment's due process protections. Grantees were given no meaningful opportunity to contest the AI's characterizations before losing funding. In several cases cited in the ruling, ChatGPT's assessments contained factual errors that a cursory human review would have caught.
The AI governance question gets real
For months, legal scholars and AI ethicists have debated hypotheticals about algorithmic decision-making in government. This ruling transforms those abstractions into precedent. The court did not hold that AI tools can never inform federal decisions—but it established that outsourcing consequential determinations to a language model, without adequate human oversight or procedural safeguards, fails constitutional muster.
The implications extend beyond DOGE. Federal agencies increasingly experiment with AI for everything from benefit eligibility screening to regulatory enforcement prioritization. McMahon's ruling suggests that any such deployment must include robust human review, clear documentation of how AI outputs are weighted, and meaningful opportunities for affected parties to challenge automated assessments.
The political fallout
DOGE was created to demonstrate that government could be leaner and smarter. Instead, its flagship initiative has produced a ruling that characterizes its methods as both legally deficient and intellectually lazy. The $100 million in grants must now be restored, and the administration faces the prospect of similar challenges to other AI-assisted cuts.
Elon Musk, who has championed DOGE's mission, has not publicly commented on the ruling. The White House issued a statement calling the decision "judicial overreach" and indicating plans to appeal.
Our take
There is nothing inherently wrong with using AI to help identify inefficiencies in government spending—the federal budget is vast, and computational tools can surface patterns humans might miss. But DOGE's approach confused the tool for the decision-maker. ChatGPT is a text-prediction engine optimized to sound plausible, not a policy analyst equipped to weigh statutory obligations, contractual commitments, and programmatic merit. Treating its outputs as self-executing judgments was not innovation; it was abdication. Judge McMahon's ruling is less a rebuke of artificial intelligence than a reminder that intelligence—artificial or otherwise—requires judgment, accountability, and the humility to know what it doesn't know.
