12APR

SAVE System Flags 1 in 6 Wrongly

3 min read

15:24UTC

DOGE's expanded voter screening system has a 17% error rate, and the Trump administration admitted it worked with True the Vote to probe voter rolls. The DOJ privacy officer resigned rather than implement the plan.

← Back to First votes exceed every forecast Jump to analysis ↓

PoliticsDeveloping

Key takeaway

The voter screening system gets it wrong one time in six, and its architects collaborated with election-denial activists.

DOGE (Department of Government Efficiency) expanded the DHS SAVE System to identify non-citizen voters. Early reports show a 17% error rate, meaning roughly one in six eligibility flags is wrong ¹. Applied at scale across state voter rolls, that rate would challenge the registration status of millions of legitimate voters.

Democracy Docket confirmed that the Trump administration admitted DOGE worked with True the Vote, an organisation involved in election-denial activism, to probe voter rolls using Social Security Administration data ². A DOGE employee signed a "voter data agreement" with True the Vote on 24 March 2025. Two former DOGE staffers were referred to a watchdog for possible Hatch Act violations over these communications.

On 3 April, the DOJ privacy officer resigned rather than implement the voter data-sharing plan between DOJ and DHS ³. A government official concluded the plan was legally or ethically indefensible and left rather than execute it. The resignation is an action, not a statement: it signals internal legal assessment that the data-sharing programme crosses a line the privacy officer was unwilling to cross. The combination of a 17% error rate, collaboration with an election-denial group, and the resignation of the official tasked with ensuring legal compliance creates a credibility problem for the verification infrastructure that the executive order and DOJ litigation are designed to build.

Deep Analysis

In plain English

The SAVE system is a federal database originally built to check whether people applying for government benefits (like Medicaid or food stamps) are US citizens or legal residents. The Trump administration expanded it to check whether registered voters are citizens. The problem is that the system was not built for voter rolls. It matches names, addresses, and dates of birth between two databases that use different formats. When names are spelled slightly differently or addresses are recorded differently, the system produces a false 'possible match' , flagging a legitimate citizen as a potential non-citizen. Early data shows this happens roughly one in six times. Separately, the administration worked with True the Vote , a group known for promoting claims that the 2020 election was stolen , to use Social Security Administration data to probe voter rolls. A DOGE employee formally signed an agreement with this group. The official whose job was to protect the privacy of government data resigned rather than implement the plan.

Deep Analysis

Root Causes

The SAVE system's 17% error rate is a structural consequence of data mismatch between federal immigration records and state voter registration databases.

SAVE was built to verify benefit applicants against USCIS (immigration) records. Voter registration data uses different name fields, address formats, date-of-birth precision, and identifier types than benefit application data. When the two databases are cross-referenced, transliteration errors, hyphenation differences, maiden-name mismatches, and address format discrepancies produce false-positive matches that cannot be resolved without individual manual review.

The True the Vote collaboration amplifies this structural problem: True the Vote's methodology in the 2022 cycle was based on commercial address-change databases that were repeatedly found to produce high false-positive rates for voters who had moved within the same county. Combining SAVE's immigration-mismatch errors with True the Vote's address-change methodology compounds the error accumulation rather than correcting it.

What could happen next?

Risk
A 17% flag error rate applied to tens of millions of voter records would generate hundreds of thousands of wrongful challenge notices, overwhelming state election administration capacity to process them before November 2026.
Precedent
The DOGE-True the Vote data agreement sets a precedent for private political organisations accessing federal SSA data for election roll challenges, creating an accountability gap between formal government authority and partisan-aligned data use.

Source Landscape

This story draws on mixed-leaning sources from United States

United States

Primary parallel: Florida's 2012 pre-election voter roll purge, ordered by Governor Rick Scott, used a state database to flag approximately 182,000 voters as potential non-citizens. Of those initially flagged, 87% were US citizens, producing an error rate structurally comparable to SAVE's 17%. Federal courts blocked the purge. The pattern , high initial error rate, federal intervention, court-ordered halt , is the closest operational precedent for the current SAVE deployment.

Counter-parallel: Georgia's 2018 voter roll maintenance programme, administered by then-Secretary of State Brian Kemp, removed approximately 340,000 voters using a database matching approach. The accuracy of the removals was disputed but the process survived federal court challenge, demonstrating that database-driven roll purges are not automatically struck down simply because of high flag error rates.

Consensus view: Government Accountability Project's Tom Devine, who specialises in federal whistleblower cases, assesses the privacy officer's resignation as an unusually significant signal: senior career officials resign over implementation rather than policy when they have assessed the plan as legally indefensible and concluded that execution exposes them to personal liability, not merely professional disagreement.

Counter-view: Cato Institute's Walter Olson argues the 17% SAVE error rate is misleading because it conflates false-positive flags with actual erroneous voter challenges , a flag triggers further review, not immediate removal, and the relevant question is the error rate at the point of final removal determination, which has not been measured.

Key tension: Whether the SAVE System's 17% flag error rate translates into an equivalent rate of wrongful registration challenges, or whether the downstream review process filters out false positives before they affect voter eligibility.

Sources:NPR·Democracy Docket

Mentions:DOGE →True the Vote →Hatch Act →Florida →Prima →Georgia (U.S. state) →DHS →Social Security Administration →Trump administration →SAVE System →Democracy Docket →SAVE Act →

First Reported In

Update #2 · First votes exceed every forecast

NPR· 12 Apr 2026

Read original →

Different Perspectives

Labour-market economists

Economists note June payrolls rose just 57,000, about half the forecast 115,000, with April and May revised down further. They call it the only development this week bearing directly on how incumbents can run on the economy in November.

Alaska political observers

The state Supreme Court's reinstatement of Dan J. Sullivan of Petersburg to the 18 August primary ballot means two men named Dan Sullivan, one the sitting senator, may both appear. Observers moved the race to Toss-up on the ballot mechanics alone, not any shift in the campaign.

OpenSecrets campaign-finance analysts

Analysts flag that all four national committees, the NRSC included, can now form joint fundraising committees combining donor money with full coordination. They expect the DSCC, NRCC and DCCC to match the move before the effect shows up in filings.

NRSC strategists

The NRSC told campaigns on 30 June to fold independent spending into fully coordinated vehicles now that the Supreme Court has struck the caps. Strategists see it as converting the RNC's roughly $110m cash edge into leverage precisely where challengers are outspending Republican incumbents.

Democracy Docket / voting-rights litigators

Litigators note DOJ is now 0-for-6 on trial losses yet still climbing the appellate ladder through the Sixth Circuit en banc bid. They read the persistence itself as the point: keep the underlying dispute alive past November regardless of the win rate.

Fulton County election officials

Fulton County had argued in court that the subpoena was meant to "target, harass and punish" perceived opponents, and the 7 July ruling ended that specific demand. Officials treat the outcome as proof the criminal track was pressure, not a genuine prosecution.