A federal judge just declared that the Department of Government Efficiency crossed constitutional lines when it deployed ChatGPT to screen and cancel over $100 million in humanities grants. In a scathing 143-page ruling released Thursday, US District Judge Colleen McMahon found that DOGE’s AI-powered process for identifying diversity, equity, and inclusion (DEI) programs amounted to unconstitutional discrimination. The decision marks one of the first major legal rebukes of government agencies using AI tools to make consequential decisions without human oversight.

OpenAI‘s ChatGPT just became exhibit A in a landmark constitutional case. The Department of Government Efficiency – the newly formed agency tasked with cutting federal spending – thought it found the perfect efficiency tool: an AI that could scan thousands of grant applications and flag anything related to diversity programs. Turns out, that was both legally disastrous and constitutionally prohibited.

US District Judge Colleen McMahon didn’t mince words in her 143-page decision released Thursday. The ruling dismantles DOGE’s entire approach to eliminating grants from the National Endowment for the Humanities, finding that the agency “used the mere presence of particular, protected characteristics to disqualify grants from continued funding.” Over $100 million in approved grants got axed in the process.

Here’s how DOGE’s AI experiment went sideways. Instead of having human experts review grant applications for actual merit or compliance issues, the agency fed descriptions into ChatGPT with instructions to identify anything touching diversity, equity, and inclusion themes. The AI flagged programs. DOGE canceled funding. No appeals process, no human review of context, no consideration of whether the flagged content was actually problematic or just mentioned demographics as part of legitimate research.

The case emerged from a 2025 lawsuit filed by humanities organizations whose grants vanished overnight. When lawyers started digging into DOGE’s process, they uncovered internal documents showing the ChatGPT workflow. The revelation that a large language model – trained on internet text and prone to inconsistent outputs – was making binding decisions about constitutional rights triggered immediate legal challenges.

Judge McMahon’s ruling goes beyond just this specific case. She establishes that using AI to screen for protected characteristics without proper safeguards violates equal protection principles, even when the agency claims efficiency as justification. “The Constitution does not have an exception for algorithmic convenience,” the decision states, creating potential precedent for how federal agencies can and cannot deploy AI tools.

The timing couldn’t be more significant for the AI industry. OpenAI has been aggressively marketing ChatGPT Enterprise to government clients, positioning the technology as a productivity multiplier for bureaucratic tasks. But this ruling exposes the gap between AI’s pattern-matching capabilities and the nuanced judgment required for decisions affecting civil rights. A chatbot that’s helpful for drafting emails becomes legally toxic when given authority over constitutional matters.

DOGE’s defense argument – that they were simply using modern tools to enforce existing anti-DEI executive orders – fell flat with the court. McMahon noted that even if eliminating DEI programs were a legitimate goal, the method matters. Using an AI to conduct what amounts to content-based discrimination without individualized review crosses the line from policy enforcement into constitutional violation.

The fallout extends beyond just reinstating the canceled grants. Federal agencies that rushed to adopt AI screening tools for everything from visa applications to contract reviews are now reassessing their legal exposure. The ruling suggests that whenever AI systems make decisions touching protected classes – race, gender, national origin – agencies need robust human oversight and clear constitutional guardrails.

For OpenAI, the case presents an uncomfortable reality check. The company has positioned ChatGPT as ready for enterprise and government deployment, but hasn’t publicly addressed liability questions when clients use the tool for high-stakes decisions. The ruling doesn’t fault OpenAI directly, but it demonstrates how their technology can amplify discriminatory outcomes when deployed without proper governance frameworks.

Legal experts are already comparing this to earlier cases where algorithmic decision-making in criminal justice and lending faced constitutional challenges. But DOGE’s ChatGPT debacle is arguably more brazen – the agency apparently made no effort to validate the AI’s classifications or build in human review checkpoints. It’s government automation at its most reckless.

The practical impact hits immediately. DOGE must now reverse the grant cancellations and establish a constitutional review process if they want to continue scrutinizing NEH funding. More broadly, every federal agency using or considering AI tools for consequential decisions just got a clear warning: efficiency gains don’t override constitutional requirements, and courts will look past the technology hype to examine actual impacts on citizens’ rights.

This ruling establishes a critical boundary for AI deployment in government operations. While agencies will continue adopting tools like ChatGPT for efficiency gains, they can’t outsource constitutional decision-making to algorithms without safeguards. For the broader AI industry, it’s a wake-up call that enterprise clients – especially government agencies – need more than just powerful models. They need governance frameworks that prevent their efficiency tools from becoming liability engines. DOGE’s experiment shows what happens when technologists forget that some decisions require human judgment, legal expertise, and constitutional awareness that no chatbot can replicate.