On January 20, 2025, President Trump signed an executive order creating the Department of Government Efficiency. DOGE, as it became known, was not a department in any traditional sense. It had no statutory foundation. Congress never created it. And yet, within weeks, it was canceling federal grants, terminating federal employees, and claiming credit for shuttering the United States Agency for International Development.
The question that has been working its way through federal courts ever since is a straightforward one is can an entity with no legal basis exercise that kind of power over the federal government?
On March 23, 2026, United States District Judge Tanya Chutkan gave a partial but important answer. She refused to dismiss the two most constitutionally significant claims against DOGE. In doing so, she rejected an argument from the government that was, to put it plainly, breathtaking in its implications.
What DOGE Promised, and What It Delivered
Elon Musk originally promised to cut one trillion dollars from the federal budget. He later revised that figure down to $150 billion. By the time he left government in May 2025, DOGE’s own website claimed savings of roughly $180 billion, represented on what the organization called a “wall of receipts.”
Analysts across the political spectrum were not convinced. The Wall Street Journal conducted its own analysis of government contract data and found that DOGE’s claimed savings of $55 billion were significantly overstated, with actual projected savings closer to $2.6 billion over the following year. Nat Malkus, a senior fellow at the conservative American Enterprise Institute, found that DOGE’s methodology was fundamentally flawed: agencies list maximum contract values the way a credit card lists a credit limit, and canceling the card does not mean you saved the full limit. AEI’s analysis put real savings closer to $10 billion. The libertarian Cato Institute similarly found that itemized, verifiable cuts were roughly half of what DOGE claimed, and that even those figures contained numerous errors and inflated values. Jessica Riedl, a budget expert at the conservative Manhattan Institute, calculated that when buyouts came off the books, the workforce savings would amount to roughly 0.1 percent of federal spending.
The errors in DOGE’s accounting were not minor. DOGE claimed it would save $740,457 by ending a lease tied to the Obama Presidential Library. But the government had already planned to end that lease, and the landlord confirmed the government was still occupying the space and paying rent. In another case, DOGE originally claimed a single contract cancellation saved $8 billion. The actual figure, after correction, was $8 million. More than a quarter of the contracts listed by DOGE had already been fully paid, meaning canceling them produced no savings at all.
The problem was what DOGE’s cuts actually cost. Some government analyses estimated that DOGE’s activities generated costs that exceeded its savings. Independent projections suggested that downstream effects, including lost productivity, rehiring costs, and disrupted programs, could ultimately burden taxpayers more than the cuts saved. The IRS, whose workforce DOGE significantly reduced, projected revenue losses from diminished enforcement capacity that dwarfed any efficiency gains. In other words, by some measures DOGE did not just fail to save money. It may have cost money.
Judge Chutkan’s ruling arrived in the middle of this story. The disruption DOGE caused was measurable. The authority it claimed was not.
ChatGPT as Grantmaker
Perhaps no single episode better illustrates how DOGE actually operated than what happened at the National Endowment for the Humanities.
In March 2025, DOGE’s Small Agencies Team met with NEH leadership. Within twenty days, DOGE had canceled more than $100 million in grants and terminated 65 percent of the agency’s employees. The process by which it decided which grants to cut has since been revealed in depositions filed as part of a lawsuit brought by the American Council of Learned Societies, the American Historical Association, and the Modern Language Association.
Two DOGE employees, Nathan Cavanaugh and Justin Fox, were responsible for identifying which grants to terminate. Neither had any background in academic research or the humanities. Their method was to feed short grant descriptions into ChatGPT with the following prompt: “Does the following relate at all to DEI? Respond factually in less than 120 characters. Begin with ‘Yes.’ or ‘No.’ followed by a brief explanation.” They did not define DEI for the chatbot. They did not consult scholars. They did not use the NEH’s peer review process. They used ChatGPT’s yes-or-no replies as the basis for a spreadsheet that determined which grants would be cut.
Grants flagged and terminated under this process included a documentary on Jewish women’s slave labor during the Holocaust, an archival project on Italian American history, a project to digitize photographs of Appalachian communities, and multiple efforts to preserve endangered Native American languages. Cavanaugh later testified that he believed a well-read person could make these calls without specialized expertise. When asked which books had informed his judgment, he acknowledged he had not consulted any.
In his deposition, Cavanaugh said the goal had been to reduce the federal deficit from two trillion dollars to close to zero. He was then asked whether DOGE had in fact reduced the federal deficit. “No, we didn’t,” he replied.
The NEH’s acting chair, Michael McDonald, later testified that in his more than two decades at the agency, only a handful of appropriated grants had ever been terminated. He also testified that he had ceded his authority over termination decisions to DOGE in writing, telling Fox that “it’s your decision on whether to discontinue funding any of the projects on this list.” He said in his deposition that he had been unaware DOGE was using ChatGPT and that he would have preserved the Holocaust-related grants had he known.
The depositions became viral online. A federal judge subsequently ordered them removed from YouTube, citing harassment concerns, over the objection of the plaintiff organizations.
The Lawsuit
A coalition of nonprofit organizations, including the Japanese American Citizens League, the Sierra Club, the Union of Concerned Scientists, and OCA, filed suit against DOGE and sixteen federal agencies. Their complaint had four claims.
One claim was ultra vires, meaning DOGE officials were acting without any lawful authority. Another charged a separation of powers violation: DOGE was canceling funds and controlling spending that Congress had appropriated, authority it was never given. A third targeted Musk personally, arguing he was exercising the powers of a principal officer without having been confirmed by the Senate, as the Appointments Clause requires. The fourth was an APA claim against the agency defendants for taking unlawful agency action.
Judge Chutkan dismissed the separation of powers and APA claims. She allowed the Appointments Clause and ultra vires claims to proceed.
The mainstream coverage focused on what survived. But the reasoning behind each decision is what matters.
The Appointments Clause Claim: The Government’s Remarkable Argument
The Appointments Clause sets out a straightforward requirement. Principal officers of the United States must be nominated by the President and confirmed by the Senate. Inferior officers can be appointed by the President alone, or by department heads, or by courts. The one thing the Clause simply does not allow is someone wielding the powers of a principal officer without any Senate confirmation at all.
The government’s response to this was striking. Its position was that DOGE’s leader falls outside the Appointments Clause entirely, because his office was never lawfully created and his authority was never lawfully conferred. The argument, stripped to its core, was that the illegality of DOGE’s creation is precisely what exempts it from constitutional oversight.
Judge Chutkan found this unpersuasive. She described the implications as “disquieting.” Taken to its logical conclusion, the government’s theory would let any President sidestep the Appointments Clause by the simple act of creating an office without congressional approval and installing someone in it without a Senate vote. The court noted that this would represent a direct assault on two of Congress’s core constitutional functions at once.
The court also examined whether DOGE’s leader qualifies as an “officer” in the constitutional sense, which requires two things: that the person occupies a continuing position established by law, and that they exercise significant authority. On both counts, the complaint held up. The position was not temporary in any meaningful sense and did not disappear when Musk left. The powers attributed to DOGE’s leader, ordering mass firings, canceling grants and contracts, shuttering entire agencies, are not the kind of incidental duties that fall below the constitutional threshold. They are, by any measure, the exercise of significant governmental authority.
The Ultra Vires Claim, Acting Without Any Authority at All
Ultra vires is a Latin phrase meaning “beyond the powers,” and in administrative law it describes exactly what it sounds like: an executive official acting without the legal authority to do so.
The government’s first response was that the nonprofits had failed to identify any specific statute DOGE had violated. There was some logic to this. Courts approach ultra vires claims carefully, and the Supreme Court’s recent decision in Nuclear Regulatory Commission v. Texas raised the bar somewhat for when such claims can proceed.
Judge Chutkan was not moved. She drew a precise distinction that the government’s argument had glossed over. The NRC case dealt with an agency that had allegedly stretched beyond a specific grant of statutory authority it actually possessed. This case was different in kind. The plaintiffs were not arguing that DOGE overstepped its powers. They were arguing that DOGE had no powers to overstep. Acting without any legal authority whatsoever is a more fundamental problem than acting in excess of authority you have, and nothing in NRC closes the door on judicial review when that is the allegation.
The government made a second attempt, pointing to 5 U.S.C. Section 3161, a provision that allows a President to create temporary organizations by executive order for specific projects of up to three years. The court was unimpressed. Canceling federal grants, ordering mass firings, and shutting down entire agencies stretches the ordinary meaning of the word “project” well past its breaking point. Congress did not hide that kind of sweeping authority in a single statutory word, and the court declined to read it that way.
What Was Dismissed, and Why
The APA claim did not fail because the court lacked jurisdiction. It failed because the complaint was written too broadly. Sixteen federal agencies, countless different actions, all folded into a single count. The APA does not work that way. It requires a plaintiff to point to something specific: a discrete agency action that caused a concrete harm. A complaint that sweeps across the entire federal government without identifying particular actions is not an APA claim. It is a grievance. The court left the door open for more targeted APA claims to be refiled, but this one could not stand as written.
The separation of powers claim ran into a different problem. The D.C. Circuit recently held in Global Health Council v. Trump that a plaintiff cannot take what is really a statutory violation and reframe it as a constitutional one. The nonprofits’ argument that DOGE violated the separation of powers was, when examined closely, an argument that DOGE violated specific statutes governing appropriations and the creation of agencies. That is a statutory claim. Labeling it constitutional does not make it so, and allowing that kind of relabeling would let litigants avoid congressional limits on judicial review whenever they found it convenient.
The Jurisdiction Questions
Two jurisdictional issues required resolution before the court could reach the substance of the case.
On the grant terminations, the Tucker Act presented a barrier. That statute gives the Court of Federal Claims exclusive authority over contract-based claims against the federal government above ten thousand dollars. A district court order reinstating a terminated grant would effectively be an order compelling the government to pay money under a contract, which is precisely what district courts cannot do under the Tucker Act. The court drew a sensible line, though. Challenging a grant termination for money is different from challenging the policy that produced it. The former belongs in the Court of Federal Claims. The latter can still be addressed here.
The mass firings raised a separate jurisdictional question involving the Civil Service Reform Act, which routes federal employment disputes through the Merit Systems Protection Board and the Federal Labor Relations Authority. The nonprofits argued they had nowhere else to go, since those forums are open to federal employees and unions, not outside organizations. The court acknowledged this but was not persuaded. When Congress builds a comprehensive system for resolving a particular category of disputes, the absence of a provision for outside parties is itself a signal that Congress did not intend them to be part of the process.
Musk Left. The Case Didn’t.
Musk’s tenure ended last May when he hit the 130-day ceiling that applies to special government employees. The government moved quickly to argue that his departure mooted the Appointments Clause claim entirely.
That argument did not hold. Federal Rule of Civil Procedure 25(d) is straightforward on this point: when a public official sued in an official capacity leaves office, the successor steps in automatically. The claim does not evaporate with the individual. It follows the office. Whoever runs DOGE today inherits the constitutional question.
There is also the matter of what Musk did while he was there. If the plaintiffs prevail on the Appointments Clause claim, the court could vacate policies and cuts that Musk put in place that are still causing harm today. The Supreme Court has suggested that actions taken by an unconstitutionally appointed officer may be void from the start. That is a live question with real consequences, and it does not become moot simply because the officer in question no longer holds the position.
What Comes Next
Before this ruling, the case had included a coalition of states as co-plaintiffs. They voluntarily dismissed their claims, leaving the nonprofit organizations to carry the litigation forward alone. The court also denied the nonprofits’ motion for expedited discovery, though without prejudice. They can refile a revised request now that the case has been narrowed.
Discovery will proceed on the two surviving claims. At the same time, the Trump administration separately asked the Supreme Court, on the very day of this ruling, to shut down a DOGE inquiry being pursued by Citizens for Responsibility and Ethics in Washington.
The Larger Question
The government’s position throughout this litigation has been, at its foundation, that DOGE’s leader cannot be subject to the Appointments Clause because his power was never lawfully granted in the first place. Judge Chutkan’s answer to that was direct: the absence of lawful authority is not a shield from constitutional scrutiny. It is a reason for more of it.
That is not a novel legal principle, but it is one worth stating plainly. The Appointments Clause exists because the founders understood what happens when individuals accumulate significant governmental power without accountability to any confirmatory process. It is not a technicality to be litigated around. It is a structural check, and it applies with particular force when the power being exercised has no clear legal basis to begin with.
Whether this case ultimately produces a ruling that reshapes how DOGE or its successor operates remains to be seen. For now, the constitutional questions are before a court, and the people asking them have standing to be heard.
Sources:
New Mexico v. Musk, No. 25-cv-429 (D.D.C. Mar. 23, 2026), Memorandum Opinion, ECF No. 119 (Judge Tanya S. Chutkan)
New Mexico v. Musk, 784 F. Supp. 3d 174 (D.D.C. 2025)
Global Health Council v. Trump, 153 F.4th 1 (D.C. Cir. 2025)
Nuclear Regulatory Commission v. Texas, 605 U.S. 665 (2025)
Department of Education v. California, 604 U.S. 650 (2025)
NIH v. American Public Health Association, 145 S. Ct. 2658 (2025)
Trump v. CASA, Inc., 606 U.S. 831 (2025)
Exec. Order No. 14,158, 90 Fed. Reg. 8,441 (Jan. 20, 2025)
Exec. Order No. 14,210, 90 Fed. Reg. 9,669 (Feb. 11, 2025)
Exec. Order No. 14,222, 90 Fed. Reg. 11,095 (Feb. 26, 2025)
Wall Street Journal, “DOGE Claims It Has Saved Billions. See Where.” (Feb. 22, 2025)
American Enterprise Institute, Nat Malkus, “How Much Money Is DOGE Saving Taxpayers?” (Apr. 17, 2025)
Cato Institute, Dominik Lett, budget policy analysis (May-June 2025), as cited by PolitiFact and AP
Manhattan Institute, Jessica Riedl, federal budget analysis (May-June 2025), as cited by PolitiFact and AP
Inside Higher Ed, “How DOGE Gutted the NEH in 22 Days” (Mar. 11, 2026)
American Council of Learned Societies, press release re: motion for summary judgment (Mar. 7, 2026)
American Council of Learned Societies v. McDonald, S.D.N.Y. (2025), Judge Colleen McMahon; Depositions of Nathan Cavanaugh (Jan. 23, 2026) and Justin Fox (Jan. 28, 2026)
Miranda Nazzaro, “Judge Refuses to Drop Lawsuit Over Musk Role as Trump Adviser,” The Hill (Mar. 24, 2026)