GAO Could Not Verify 96 Percent of DOGE's Claimed Grant Savings. If Your Award Is on the Wall of Receipts, the Number Next to It Is Probably Not a Number.

There is a difference between money saved and money claimed. The Government Accountability Office spent the better part of a year trying to measure the gap, and on August 6, 2026 it published what it found.

GAO-26-108615, "DOGE Wall of Receipts: More Transparency Needed on How Savings Are Derived from Contract, Grant, and Lease Terminations," is not a rhetorical document. It is a methodology audit. GAO took the public savings ledger the U.S. DOGE Service has maintained since February 17, 2025, selected entries, and asked a narrow question: can the reported savings figure be traced to a documented calculation?

For grants, the answer was no in 96 percent of cases. DOGE did not provide sufficient information to verify the method used to calculate 96 percent of DOGE-reported grant savings.

That single figure deserves careful handling, because it is easy to overstate and easy to dismiss. It does not say the terminations did not happen. It says the dollar amounts attached to them cannot be checked.

What the report examined

As of July 7, 2026, the Wall of Receipts reported approximately $110 billion in savings across three categories: contracts, grants, and real estate leases. GAO reviewed how those figures were derived against DOGE's own publicly stated methodology.

The findings, by category:

The contract example GAO singled out is the clearest illustration of the problem. DOGE claimed $1.7 billion in savings on a Defense Health Agency IT services contract. GAO found the contract had never been terminated or modified. No savings were achieved. The entry was not an estimate that turned out high — it described an event that did not occur.

The lease finding is a subtler version of the same failure. A termination already in the pipeline before an office existed is not a saving attributable to that office. Counting it is not fraud; it is an attribution error. But it is an attribution error repeated across 108 of 264 entries, which stops being a rounding issue and starts being a methodology.

GAO's recommendation was modest and specific: DOGE should prominently display known data quality issues and limitations on the Wall of Receipts. GAO noted that the site launched in February 2025 and, through July 7, 2026, had not been updated to address methodology or data limitations. The U.S. DOGE Service did not respond to the report.

Why the grants number is structurally different from the contracts number

The contracts finding is about misapplied methodology. The grants finding is about missing documentation — GAO could not see the arithmetic at all. That distinction matters for anyone whose award appears on the ledger.

Grant savings are genuinely harder to compute than contract savings, and the difficulty is not an excuse but it is a real feature of the mechanics. When a discretionary grant is terminated mid-period-of-performance, the "saving" is the difference between the total obligated amount and what the recipient has already drawn down and validly incurred, minus allowable closeout costs, minus any unliquidated obligations the government still owes for costs incurred before the termination effective date. That figure is not knowable on the day of termination. It is knowable after final financial reporting, which arrives 120 days after the end of the period of performance under 2 CFR 200.344.

A savings number published on the day of termination is therefore necessarily an estimate — of a quantity that will be settled later, by the recipient's final SF-425, in an amount that is often materially different. Multi-year awards make this worse: a five-year award terminated in year two may show the full remaining out-year obligation as "saved," even though out-year amounts were never obligated in the first place and existed only as an expression of intent in the Notice of Award.

None of that is exotic. It is ordinary federal financial assistance accounting. What GAO found is that for 96 percent of grant entries, there was no documentation showing which of these conventions was applied.

What this means if your award is on the ledger

Terminated grantees have been treating the Wall of Receipts entry for their award as an authoritative government statement. It is more useful to treat it as an unsourced assertion — one that is now, as of August 6, formally characterized by GAO as unverifiable in the overwhelming majority of grant cases.

Four practical implications:

1. The number attached to your award is not a determination of what you owe or what you were denied. Your financial obligations after a termination are fixed by the termination notice, the terms of your award, and 2 CFR 200.343–200.345 — not by a public dashboard. If an entry shows a savings figure larger than your remaining unobligated balance, that discrepancy is evidence of a methodology error, not a debt.

2. Your own record is now the better record. GAO's finding establishes that the government's public ledger lacks documented derivation for nearly all grant entries. Your award file — the Notice of Award, obligation history in the payment system, drawdown records, incurred-cost documentation, and final financial report — is a documented, auditable account of the same transaction. Where the two disagree, the burden of explanation has shifted meaningfully.

3. In litigation and appeal, the report is usable context, not a merits argument. It does not establish that any particular termination was unlawful. What it does establish is that the public accounting of the termination program is not reliable, which is relevant to any claim that turns on the government's stated rationale or the fiscal justification offered for a termination decision. Counsel handling matters under the agency-priorities clause and termination-for-convenience authority — the terrain covered in our analysis of the Talwani ruling and AHRQ clawbacks — should read GAO-26-108615 in full rather than the headline.

4. Reconstruct now, not later. If your award was terminated in 2025 or 2026, assemble a single-page reconciliation: total obligated, total drawn, costs incurred and unreimbursed, closeout costs claimed, and the actual de-obligated amount as reflected in your final report. That document is what you will need if the ledger figure is ever cited against you, in a subsequent competition, in a single audit, or in a congressional inquiry.

The oversight context this lands in

GAO-26-108615 is not an isolated finding. It arrives alongside two other 2026 GAO products that describe the same underlying condition — a federal grants enterprise whose money is easier to move than to track.

GAO-26-109100, "Combating Fraud: Managing Risks in Federally Funded, State-Administered Programs," issued July 23 and reissued with revisions August 7, 2026, examined 20 large state-administered programs accounting for roughly $1.1 trillion in federal obligations in FY 2025. GAO found that only 5 of the 20 had documented evidence consistent with identifying fraud risks and assessing their likelihood in order to prioritize action. The other 15 did not. GAO carries 22 open recommendations to agencies on fraud risk management, and estimates government-wide fraud losses at $233 billion to $521 billion annually — 3 to 7 percent of federal obligations — based on FY 2018–2022 data.

GAO-26-108283, "Grants Management: Efforts to Address Challenges Through Government-wide Collaboration," issued April 14, 2026, documented approximately $1.2 trillion in grants to tribal, state, local, and territorial entities in FY 2024 and the substantial variation in how programs are designed and administered across agencies. It also described the Council on Federal Financial Assistance, established by OMB in August 2023, which represented 38 of 40 grantmaking agencies and over 99 percent of federal grant funding obligated in FY 2024.

Read together, the three reports describe a system in which $1.2 trillion moves annually, most large state-administered programs have not documented a fraud risk assessment, and the highest-profile public accounting of savings from cutting that system cannot be verified in 96 percent of grant cases. That is the same structural problem POGO identified in its finding that a large share of federal grant spending is untraceable past the first recipient.

The uncomfortable synthesis: the data infrastructure is not good enough to support confident claims in either direction. It cannot reliably substantiate the scale of waste, and it cannot reliably substantiate the scale of savings from eliminating it.

What changes for grant seekers

Not much operationally — and that is the point worth making clearly, because the temptation after a report like this is to over-read it.

Your competitive strategy does not change. Your compliance obligations do not change. The Wall of Receipts is not an eligibility screen, and no reviewer is consulting it.

What does change is how you should weigh public savings claims when they appear in the justification for a program's elimination or an agency's budget. A savings figure that GAO has characterized as lacking documented derivation is not a fact you need to accept as the premise of a planning conversation with your board.

Three concrete adjustments:

If you are writing a case for support that references federal funding volatility, cite the GAO report number rather than press coverage. Boards respond differently to a numbered audit than to a news headline, and GAO-26-108615 is a citable primary source.

If you are a subrecipient, confirm your pass-through entity's understanding of your award's status directly, in writing. Public dashboards and agency press releases have not been reliable indicators of what was actually de-obligated.

If you are budgeting against a terminated award, budget against your final financial report, not against any public figure. The reconciliation you build for defensive purposes is the same document that makes your next single audit straightforward.

The recommendation nobody acted on

GAO asked for one thing: display the known limitations. Label the estimates as estimates. Note where a figure is a projection of out-year amounts that were never obligated. Flag entries where the underlying action was already in progress.

That is a disclosure request, not a substantive challenge to any termination. It is the kind of recommendation agencies typically accept without argument, because accepting it costs nothing and refusing it looks worse than the underlying error.

The U.S. DOGE Service did not respond.

For grantees, the operative lesson is the durable one: when the public record of a decision affecting your organization cannot be verified, your own documentation becomes the authoritative account. Build it while the files are fresh. The organizations that came through the 2025 termination wave in the strongest position were not the ones that argued loudest — they were the ones described in our termination survival guide who could produce, on request, a complete and reconciled financial record of exactly what they had been awarded, what they had spent, and what was left.