There Is No National Voter File

Be precise about how that safeguard got there, because it is not a monument. Congress has never voted a national voter list down; it has simply never voted one up. The Help America Vote Act went as far in 2002 as requiring every state to maintain a centralized statewide list, and pointedly stopped there. The fragmentation is protected by an absence, not a refusal — a federalism default nobody bothered to override, and defaults are cheap to route around, because there is no vote on record to contradict.

Immigration and Customs Enforcement is routing around it now.

Homeland Security Investigations has issued a Request for Information seeking a contractor for the "handling and secure delivery of publicly available voter registration files and voter history files to support Homeland Security Investigations (HSI) fraud detection and data segmentation activities." The winning vendor must be able to reach all fifty states and the territories, though the initial jurisdiction list is expected to be short.

Take the caveat seriously before the criticism: an RFI is market research. No contract has been awarded, no vendor selected. Agencies publish RFIs for things they never buy. This is a stated intention, not an accomplished fact.

Now read the sentence again, because the sentence is the story.

Fraud detection is the announced purpose, and announced purposes are cheap. Noncitizen voting is among the most thoroughly audited rare events in American public life, and the predicate for looking anyway has been built in public for a year — DHS pushing its SAVE citizenship-verification system at local officials, the Justice Department in court against states that won't surrender their rolls. That half is at least legible.

Data segmentation is the other half, and nobody announced it. Grant the boring reading first, because it is probably the right one: segmentation is a plain data-engineering term for partitioning a dataset — by jurisdiction, by field — for processing and delivery, which is exactly what a fifty-state normalization job consists of. The concession costs nothing. Fraud detection is per-case work, and a fraud case has a name in it. Segmentation, in either sense, is per-population work, where the deliverable is a sorted set rather than a suspect. The two halves of that sentence do not describe the same job, and the benign reading only makes the mismatch mundane rather than sinister — which is the likelier way a thing like this gets built.

Then there is the contractor, though not for the reason it first appears. ICE does not need a vendor to download public files; the files are public, that's the premise. Anyone can buy Ohio's voter roll. What a contractor supplies is not access but aggregation: fifty-plus jurisdictions in fifty-plus incompatible formats, normalized into one queryable national file.

That file exists. It has existed for roughly twenty years. L2, Catalist, TargetSmart, i360, Aristotle — a mature commercial market in exactly this product, covering all fifty states and the District of Columbia, vote history attached, records going back two decades, consumer data appended on top of the public fields. Campaigns in both parties buy it as routine. The national voter file Congress never authorized was assembled anyway, privately, and put up for sale.

Which relocates the story rather than softening it. ICE is not commissioning an unprecedented instrument; it is going to market for one already on the shelf. That is the data-broker loophole — government buying what it cannot compel — and the agency has the habit: $125 million to Thomson Reuters for personal data used in "voter fraud" work, a separate LexisNexis arrangement feeding Palantir. The safeguard in the first paragraph was defeated commercially years before an enforcement agency thought to go shopping, and nobody voted on that either. There was never anything to vote on. It happened one state file at a time, lawfully, as a service to campaigns.

The bind underneath does not resolve. Voter files are public by design, and the design is correct. The alternative — a secret list of who may vote, purged out of view — is the specific historical failure the transparency requirement was written against. You can audit a public roll; you cannot audit one you can't see. The openness that makes democratic accountability possible is the same openness that makes national aggregation a business, and it is load-bearing in both directions. Nothing here is a leak or a breach; nothing is being stolen. The commons is functioning exactly as specified, and what it was specified for turns out to be indistinguishable, at the level of the data, from letting anyone assemble the list and sell it to whoever asks.

The file is built. It was always going to be bought by someone. The only thing this RFI decides is who the customer is, and that gets settled by a contracting officer, on a schedule, in a document nobody will read.