The Department of Homeland Security has alleged that nearly 250,000 possible noncitizens may be registered to vote across four states, but election officials say the federal government has not supplied a reliable method for identifying them. That collision between a large national claim and incomplete state-level evidence has become a central feature of President Donald Trump’s election message as the November 3 midterms approach. A new AP review published Sunday found that the administration is pairing the disputed numbers with criminal enforcement, legislation and demands for tighter control of mail voting.

The available record does establish that noncitizens sometimes enter voter-registration systems. New Jersey disclosed a software failure that registered about 6,600 people who had said they were not citizens, and federal prosecutors charged a lawful permanent resident in California with falsely registering. Yet those incidents do not validate the broader federal estimates. They instead illustrate the harder administrative question: how to find genuine errors without treating imperfect database matches as proof of illegal voting.

That distinction matters because registration, ballot receipt and voting are different events. Federal law already makes voting by a noncitizen in a federal election illegal, with narrow exceptions, and a false citizenship claim on a registration form can bring additional penalties. The policy dispute is therefore not whether citizenship should be required. It is whether the administration’s evidence measures violations accurately and whether its proposed remedy could exclude eligible citizens who cannot quickly produce the specified documents.

A Federal Claim Meets State Resistance

Homeland Security Secretary Markwayne Mullin sent letters in July to California, Nevada, New Jersey and Pennsylvania warning that federal comparisons had identified tens of thousands of possible noncitizens on their rolls. The DHS release described election security as a national-security issue and asked the states to cooperate. State officials, however, said the department had not provided enough information to reproduce or assess the matches.

Pennsylvania’s top election official, Republican Secretary of the Commonwealth Al Schmidt, said DHS representatives acknowledged that they did not have high confidence that the 14,576 people named in an initial federal letter were noncitizens improperly registered. According to local reporting, federal officials promised to provide their methodology, which Pennsylvania was still awaiting. California and Nevada officials raised similar concerns about data sources and the risk of misidentifying naturalized citizens or people with outdated immigration records.

The Census Bureau added a separate and politically potent figure on August 18. Its preliminary analysis said more than 24,000 of nearly 160 million voter records from 2020 represented noncitizens at the time of that election, while more than 32 million records remained unanalyzed. The brief called its work a beginning analysis but omitted the error rates and validation details normally needed to judge whether record linkage had confused people, addresses or citizenship status. Trump cited the figure as evidence that the 2020 result was corrupted, although 24,000 ballots would not approach his national popular-vote deficit and the report did not show that its matches were confirmed cases.

Isolated Failures Are Real but Uneven

New Jersey offers the clearest evidence that administrative mistakes can put ineligible people on a roll. Governor Mikie Sherrill said a vendor’s software registered about 6,600 people between June 2023 and June 2024 even though they answered “no” when asked whether they were citizens. The state’s preliminary review found that fewer than 400 of those newly registered people voted. Officials removed affected registrations, began replacing the vendor and said there was no evidence the votes changed any election outcome.

The episode is consequential, but it points to a system defect rather than a coordinated scheme. The affected people did not attest to citizenship during the motor-vehicle transaction, according to New Jersey’s explanation, and the software transmitted them anyway. That difference can determine criminal intent as well as the appropriate cure. It also shows why registration totals alone cannot establish how many people knowingly violated the law or cast ballots.

The California prosecution is narrower. The Justice Department said Darwin Jonathan Rivera Flores, a 30-year-old Honduran national and lawful permanent resident, checked a box stating that he was a citizen on an online registration. The criminal complaint alleges false citizenship and fraudulent registration, each carrying a five-year statutory maximum. Prosecutors also stated that Rivera Flores never voted. Those charges remain allegations, and he is presumed innocent unless proved guilty.

What the Numbers Actually Measure

Database matching can be useful for maintaining voter rolls, but its accuracy depends on the fields, update schedules and thresholds used. Immigration records may predate naturalization. Commercial voter files can associate a person with an old address, while common names and incomplete identifiers can generate false matches. Without a published denominator, false-positive rate and manual validation sample, a list of “possible” noncitizens cannot be converted into a count of illegal registrants.

The enforcement baseline is much smaller than the headline claims. A Reuters investigation found 129 federal prosecutions under the voting-by-aliens statute enacted in 1996. That total cannot capture every violation, because detection and referral vary, but it provides no evidence of hundreds of thousands of prosecuted votes. The governing federal statute also underscores the legal risk: noncitizens who vote can face criminal exposure and, under immigration law, deportation.

The evidence therefore supports two findings at once. Improper registrations occur and deserve prompt correction, transparent investigation and, when intent can be proved, prosecution. But the public data released so far do not substantiate the administration’s largest estimates or demonstrate that noncitizen voting has occurred at a scale capable of deciding federal elections. Treating those propositions as interchangeable obscures the gap between an administrative flag and a verified unlawful ballot.

The Legislative Stakes

Trump is using the claims to press the Senate to pass the Safeguard American Voter Eligibility Act, or SAVE America Act. The House-passed legislation would require documentary proof of citizenship for federal voter registration and photo identification at the polls, while creating alternative procedures for some applicants. Republicans supporting the measure say a uniform federal standard would improve confidence and stop ineligible registration before it happens. The bill remains stalled because it lacks the votes needed to overcome a Senate filibuster.

Opponents focus on the population of eligible citizens who may struggle to comply. A national survey conducted by the University of Maryland’s Center for Democracy and Civic Engagement and partner organizations estimated that 21.3 million voting-age citizens lacked ready access to documentary proof of citizenship; 3.8 million said they had none. The study measures document access, not certain disenfranchisement, because people may obtain records or use an alternative process. It nevertheless identifies the potential administrative burden if new requirements take effect close to an election.

Courts have also limited efforts to add documentary requirements through state or executive action. A federal judge this week blocked an Ohio rule at motor-vehicle offices, finding that written citizenship attestation satisfied the National Voter Registration Act for the challenged applications. Ohio officials argue that documents are a commonsense safeguard and may appeal. The ruling does not resolve Congress’s power to rewrite federal registration law, but it shows why the outcome in the Senate would materially change the legal landscape.

The November Test

Election officials now face two simultaneous credibility tests. They must detect and correct real errors, including the kind New Jersey documented, while demanding enough information to validate federal match lists before removing anyone. A transparent process would disclose methods, permit affected voters to respond and separate registration problems from proven ballots. Those safeguards protect both election integrity and eligible voters; neither requires accepting unverified numbers.

The next evidence to watch is concrete: whether DHS publishes its matching methodology, how many flagged records states confirm after individual review, whether prosecutions establish actual voting, and whether Congress changes the governing rules before November. For now, the record demonstrates isolated violations and administrative failures, not the scale asserted in federal rhetoric. The political consequences may nonetheless be large, because disputed claims can shape confidence in an election long before any ballot is counted.