A software failure that improperly registered thousands of noncitizens in New Jersey has intensified a national argument over voter rolls, citizenship verification and the limits of federal election oversight. Yet the emerging evidence also shows why confirmed illegal votes, inaccurate registrations and preliminary database matches must never be treated as the same thing.
As the 2026 midterm elections approach, we are confronting two realities at once. Administrative failures have placed ineligible people on voter rolls, and some of those individuals appear to have cast ballots. At the same time, federal officials have promoted much larger figures that remain preliminary, disputed and unsupported by a publicly available methodology.
President Donald Trump has used the controversy to press Congress to approve the SAVE America Act, which would establish new national requirements for proving citizenship and verifying voter eligibility. His administration has portrayed weaknesses in registration databases and election technology as threats aimed at the “very heart of our democracy.”
The central question is no longer whether voter registration systems can make mistakes. New Jersey has confirmed that they can. The harder question is whether the evidence demonstrates a limited but serious administrative problem or a coordinated national pattern capable of affecting election outcomes.
New Jersey’s 6,600 Voter Registration Error Changed the National Debate
The most concrete evidence comes from New Jersey, where Gov. Mikie Sherrill announced on July 21, 2026, that a software error had improperly registered roughly 6,600 people who had indicated they were not United States citizens.
According to Sherrill, the affected applicants answered “no” when a Motor Vehicle Commission keypad asked whether they were citizens. The system nevertheless transmitted their information for voter registration between June 2023 and June 2024. The governor said the registrations occurred before she entered office and blamed what she called a “serious software error.”
New Jersey’s preliminary review found that fewer than 400 people who were registered for the first time through the error later voted. Those registrations included Democrats, Republicans and unaffiliated voters distributed across the state, undermining assumptions that the incident benefited only one political party.
We must still treat the number of ballots cautiously. “Fewer than 400” describes a preliminary finding, not a completed determination that every person involved remained a noncitizen when voting or knowingly violated election law.
A follow-up statement from Sherrill’s office demonstrated why that distinction matters. By July 28, New Jersey had begun removing more than 4,500 people who had been registered solely because of the software error. Officials sent the remaining records to county election administrators for additional review because the state’s initial investigation indicated that some affected people might now be citizens or might otherwise have been legally registered.
That update significantly changes how we should describe the case. New Jersey did not simply discover 6,600 confirmed illegal voters. It discovered 6,600 improperly processed registrations, began removing thousands of clearly affected records and reserved others for individual verification.
The distinction does not excuse the original failure. A system that registers someone after that person explicitly identifies as a noncitizen has failed at one of its most basic tasks. It also exposes the individual to confusion, investigation and potentially severe immigration consequences for an error the applicant may not have caused.
The White House quickly cited the New Jersey breakdown as evidence for the SAVE America Act. It argued that the incident resulted from inadequate citizenship verification and said every unlawful ballot weakens the voice of a lawful voter.
State Reviews Show Repeated Errors but Widely Different Numbers
New Jersey is not the first state to discover apparent noncitizens in voter registration or voting records. Public reviews stretching back more than two decades reveal cases involving software problems, automatic registration systems, incomplete citizenship information, outdated immigration records and applicants who mistakenly believed they could vote.
The evidence supports concern about administrative weaknesses. It does not automatically support claims of a coordinated national operation.
Michigan conducted a statewide review after the 2024 presidential election by comparing more than 7.9 million active driving records with over 7.2 million active voter registrations. The state identified 15 people who appeared to be noncitizens and had cast ballots. When combined with a previously identified Chinese national accused of voting, the total represented approximately 0.00028 percent of more than 5.7 million ballots casted.
Michigan referred 13 cases for possible prosecution, continued investigating one and reported that one affected individual had died. Secretary of State Jocelyn Benson described the issue as serious but said officials should address it “with a scalpel, not a sledgehammer.”
Ohio reported larger numbers during a 2024 citizenship review. Secretary of State Frank LaRose said his office had identified 597 registrations associated with apparent noncitizens and found that 138 of those individuals appeared to have voted while state and federal records listed them as lacking citizenship. The state referred the evidence to the attorney general for further investigation and possible prosecution.
Ohio reached those figures by comparing voter records with information from the Bureau of Motor Vehicles, the Department of Homeland Security’s SAVE database, Social Security records and other sources. Its removal procedure required notice and an opportunity for affected people to provide proof of citizenship
These cases demo.etimes identify ballots associated with those records. Third, an initial database match does not establish a final citizenship determination or prove criminal intent.
Michigan illustrated the danger of treating every flag as conclusive in January 2026. State officials reviewed a local allegation involving 15 supposed noncitizens and found that three were citizens, four were apparent noncitizens, four had already been removed and four required more investigation. One lawful citizen had reportedly been placed under criminal scrutiny after legally voting
The lesson is straightforward. Data matching can locate records worth reviewing, but it cannot replace record-by-record investigation.
DHS Claims of More Than 250,000 Potential Noncitizens Remain Preliminary
The Trump administration dramatically expanded the controversy on July 17, when the Department of Homeland Security announced that preliminary reviews had identified more than 250,000 potential noncitizens on voter rolls in four states.
DHS listed as many as 190,832 in California, 35,152 in New Jersey, 15,903 in Nevada and 14,576 in Pennsylvania. Homeland Security Secretary Markwayne Mullin declared that “only Americans should be electing American leaders” and asked the states to cooperate with federal verification efforts.
The department’s written release used qualifying language. It described the findings as preliminary, referred to “potential non-citizens” and said there “may be as many as” the listed figures. Those qualifications are crucial because political speeches and headlines have sometimes presented the same numbers as confirmed totals.
The New Jersey evidence shows why we cannot erase those caveats. DHS initially listed 35,152 potential noncitizens in the state. New Jersey separately confirmed a specific software incident involving about 6,600 registrations, later saying more than 4,500 could be removed immediately while the rest required additional review. The two figures measure different things and should not be combined or treated as interchangeable.
NPR reported on July 31 that state officials were still seeking information about how DHS developed its national figures. Letters obtained by the outlet reportedly described 118,003 records across the four states that matched a noncitizen record using a name, date of birth, address and Social Security number. Even those stronger matches required state cooperation to conclusively establish identity and current citizenship.
A person can appear in an immigration database as a former noncitizen and later become naturalized. Records can also contain outdated addresses, variations in names, incorrect dates, or information belonging to different people. A match can therefore indicate a problem without proving one.
We should separate the data into three categories:
Potential matches: Records flagged because voter information resembles a federal or state noncitizen record.
Confirmed ineligible registrations: People whose identity and current lack of citizenship have been verified.
Confirmed unlawful votes: Verified noncitizens shown to have cast ballots in an election where they were not eligible.
Public debate often jumps directly from the first category to the third. That leap produces dramatic numbers but weak analysis.
A Reuters examination of federal court records found that 129 people had been charged under the federal noncitizen-voting law since its enactment in 1996, with 73 convicted or entering guilty pleas. The investigation found no evidence of coordinated nationwide fraud and noted that several cases involved legal residents who appeared confused about their eligibility or had received incorrect information.
That record does not mean election administrators should ignore illegal voting. It means we should describe its scale using confirmed evidence rather than projections.
Federal Law Prohibits Noncitizen Voting but Does Not Produce One National Count
Federal law already makes it illegal for a noncitizen to vote in an election held partly or entirely to choose a president, vice president, member of Congress or other federal officeholder. A violation can bring a fine, up to one year in prison, or both. The statute contains a narrow exception for certain people who reasonably believed they were citizens because of their parents and childhood circumstances.
Federal law also prohibits knowingly making a false claim of citizenship to register or vote. The Department of Justice has repeatedly stated that state motor vehicle agencies should not offer federal voter registration to applicants known to be noncitizens.
The reporting system is less uniform than the criminal prohibition.
The National Voter Registration Act requires states to maintain accurate and current voter rolls, notify applicants of eligibility requirements and impose penalties for false registration information. Its list-maintenance provisions explicitly address deaths and changes of residence, while also requiring programs to operate uniformly and without discrimination.
However, the law does not create a single mandatory nationwide report listing every suspected noncitizen registration, every confirmed removal and every associated ballot. States use different databases, definitions and investigative standards.
The Election Assistance Commission conducts the Election Administration and Voting Survey, the country’s broadest collection of state and local election data. The survey covers registration, list maintenance, ballots and election administration, but the commission warns that state collection practices vary and can limit direct comparisons.
As a result, one state might announce every record produced by an initial database screen. Another might report only individuals removed after a completed investigation. A third might disclose only cases referred for prosecution.
Comparing those numbers without explaining the methodology can create a false impression that one state has a larger problem when it may simply have a broader reporting practice.
How the SAVE America Act Would Change Voter Registration
The SAVE America Act would impose national citizenship-documentation requirements for federal voter registration.
The bill would require applicants to present documentary proof of United States citizenship, such as a passport, qualifying birth record, citizenship-indicating REAL ID or another accepted government document. It would also require states to establish procedures for applicants who lack standard documents or whose names differ across records.
The House-approved version would direct states to submit voter registration lists to DHS for comparison through the SAVE program. States would then investigate identified records, notify affected individuals, and allow them to prove citizenship before removal.
Supporters argue that the New Jersey malfunction shows why citizenship should be verified before an applicant reaches the voter roll. Under that view, signed attestations and software filters cannot provide sufficient protection when a technical failure can override a direct noncitizen response.
Opponents argue that documentary requirements could create obstacles for eligible citizens who lack passports, cannot easily obtain birth records or have changed their names through marriage, adoption or court proceedings. They also warn that federal databases can contain incomplete or outdated citizenship information.
A Reliable Citizenship Verification Process MustProtect Both Sides
An effective voter-roll system must stop ineligible registrations before they occur, identify errors quickly and provide a fair correction process.
The New Jersey incident shows what happens when a voter registration program accepts information that its own system should reject. Michigan’s review shows how a careful audit can identify a small number of potentially unlawful ballots without portraying millions of lawful votes as suspect. Ohio’s process shows why officials need several databases, direct notice and an opportunity to contest removal.
