Up to $125M for ICE to Hunt 'Voter Fraud'

Introduction

On April 27, Thomson Reuters told the public that "our $22.8 million DHS contract does not include CLEAR." Eleven weeks later, on July 14, the government published its case for renewing that same contract line. The successor puts CLEAR at the front of the bundle, priced at up to $25 million a year for up to five years. If you're registered to vote, you sit inside the population the new justification describes: it says the data has to be "readily accessible to support the presidential mandate of the identification of Voters fraud, Immigration Fraud, and National Security," per WIRED's transcription of the document.

404 Media runs that rate out to a ceiling of up to $125 million. The contract it would replace obligated a fraction of that across its entire five-year life. CLEAR is where the matching material lives: Social Security numbers, credit header data, cell phone data, utility records, license plate reads, on people who have never been charged with anything. This story is developing. Details may change.

What Thomson Reuters Published on April 27

Thomson Reuters put up a post that same day titled "Setting the Record Straight on Thomson Reuters CLEAR." Most of it corrects press coverage that had, the company said, "confused CLEAR and License Plate Recognition (LPR), a Motorola product that we offer. These are two separate products." Then comes the proof the company picked to settle it: "For example, our $22.8 million DHS contract does not include CLEAR."

That $22.8 million contract is LEIDS-5, PIID 70CMSD21C00000002, a law-enforcement database subscription running to Thomson Reuters Special Services in McLean, Virginia. USAspending now carries it at $24.0 million obligated, with performance pushed from May 31 to August 31, 2026, matching WIRED's description of the new deal's "previous equivalent contract": $24 million over five years, originally due to expire in May and since extended to August. WIRED never prints a PIID, so the succession is a well-supported inference rather than a confirmed mapping.

The replacement isn't one product either. WIRED reports it bundles CLEAR with CABS (a "Continuous Alerting Batch Solution" pulling incarceration and law-enforcement contact records with real-time location alerts), Westlaw, real-time arrest records, and a tool called TEA that feeds a risk platform named RAPID. The narrow contract Thomson Reuters held up in April to prove CLEAR wasn't involved is being replaced by a broad one where CLEAR leads the list. We covered the employee revolt over that renewal, and the organizer's firing, in a June 1 piece; those retaliation allegations are unproven and Thomson Reuters disputes them.

Inside DHS's Case for Skipping the Bidding

Sole-source procurement means no competitive bid, and the government has to write down why nobody else can do the job. Per WIRED, DHS's justification calls Thomson Reuters Special Services "the only contractor" able to deliver "continuous monitoring of up to one million individuals and entities," with "event-driven monitoring," "real-time alerts," and "model-based risk scoring."

The justification itself isn't public. A narrower, adjacent one is: DHS's April 2026 sole-source justification for renewing CLEAR through West Publishing, dollar figures redacted, stated purpose "criminal enterprise, terrorism, and immigration fraud," with no voter language anywhere. What it does supply is DHS naming its rejected alternatives out loud: "Other Law Enforcement databases such as LexisNexis and TransUnion do not offer these specific interface search capabilities." That document also inventories ten "exclusive features" of CLEAR, including cell phone, utility, and credit-bureau data.

DHS has made that same no-substitute argument since at least 2017, when an ICE/HSI limited source justification for a $13 million West Publishing contract described CLEAR supporting "Batch Requests where multiple social security numbers (SSN) and or phone numbers may be queried at one time" for benefit-fraud work. Bulk SSN matching against a name list is exactly the operation a voter-roll screen runs on, and the capability has been sitting in the contract file since 2015.

One phrase WIRED pulled from the new justification stands out: "academic risk flagging," which the document never explains. The same document authorizes identifying anyone tied to "any type of fraud of government funds," and tasks the bundle with helping identify unaccompanied minors, normally HHS's job rather than ICE's. DHS, ICE, and HHS did not respond to WIRED's questions, and there is still no government defense of the voter-fraud language beyond the sentence in the procurement file.

Reuters Checked DHS's Claims About One Deportee

The machine this data would feed is already running. Reuters reporter Peter Eisler went through federal court records and a DHS FOIA production and found 129 people charged under the 1996 noncitizen-voting statute, 86 of them during Trump's presidency (syndicated text via Business Day). DHS agents have opened voter-fraud investigations into more than 325 people across at least six states, and in none of the cases did prosecutors allege a coordinated effort. Most traced back to voter confusion or administrative mistakes.

Then there's Dora Damatta-Rodriguez, whom DHS publicly described as a "criminal illegal alien" with a record it said included alien smuggling. Reuters checked: she's been a lawful permanent resident since 1973, through marriage, with no convictions or prosecutions for the offenses DHS named, only three decades-old misdemeanors the sentencing judge ruled too old to weigh. She was deported in June 2026 over a 2016 vote she believed was legal. That reporting came from Reuters, Thomson Reuters' editorially independent news arm.

Who Benefits

DHS gets power, specifically the kind that skips a step. A sole-source award avoids the price competition and public comparison a full-and-open procurement forces into daylight, and writing "Voters fraud" into the justification converts a contested political claim into a documented contract requirement, with the data arriving through a vendor relationship instead of a warrant. On July 17, DHS Secretary Markwayne Mullin told anyone voting illegally, "we will hunt you down, we will find you and we will prosecute you." David Becker of the Center for Election Innovation and Research told the AP that 15 courts, six with Trump-appointed judges, have held the federal government cannot legally demand states' voter data: "What he's suggesting is illegal."

Thomson Reuters gets money and a government-issued endorsement of its market position. Up to $25 million a year against roughly $4.8 million actually obligated per year on the predecessor is a steep step up in rate, though a ceiling and real spending are different measures, and sole-source ceilings routinely go unspent. Against Q1 2026 revenue of $2.09 billion and $605 million returned to shareholders, even the full five-year ceiling is small. The stock is where the incentive shows up: Wells Fargo cut its price target on AI-competition concerns after the stock fell nearly 58% year over year. A company watching AI eat its legal-research franchise has a real reason to defend a federal revenue line rather than quietly drop it.

The company disputes the premise. It rejected the shareholder resolution's core claim that its products are integral to mass deportation: "We wholly dispute that claim. These allegations are without factual basis." At the June 10 annual meeting, Reuters reported chairman David Thomson saying over 95% of votes went against a human-rights review and over 3% for it, with ISS and Glass Lewis, the dominant proxy advisers, recommending a no vote. Thomson Reuters completed a second human rights impact assessment in 2025 with an outside consultancy, and spokesperson Kat Hanley called another one "duplicative and an inefficient use of resources."

The April post also says immigration status is not a search field in CLEAR, which holds on its own terms since ICE can bring status from its own systems. Nobody has explained how a database with no status field sits inside a contract for identifying immigration and voter fraud.

7.2 Million Flagged, Four Ever Charged

Database matching in service of a voter-fraud narrative has a paper trail. The Interstate Crosscheck program, run from Kansas between 2005 and 2019 under Kris Kobach, matched voters on first name, last name, and date of birth, and flagged 7.2 million "potential double registrants" across 28 states, per Rolling Stone's analysis. Four people were ever charged with double voting or deliberate double registration.

A Stanford/Harvard/Penn/Microsoft study entered into ACLU litigation calculated that following Crosscheck's standard procedure would wrongly strip more than 300 legitimate voters for every potential double vote prevented. Because common surnames cluster by demographic, so did the flags: one in six Latino voters and one in nine Black voters in the states examined, per analyst Mark Swedlund's review cited by the Brennan Center. Courts blocked Crosscheck purges in 2018; the Seventh Circuit affirmed.

That history is why a detail from Trump's July 16 primetime address matters. He said DHS had "identified approximately 278,000 noncitizens who are registered to vote in federal elections." FactCheck.org reports no released methodology, an underlying document covering only four states, and fewer than 100 noncitizens convicted of illegally voting or registering since 1982 in the Heritage Foundation's own database. Neither Trump nor DHS claimed any of them actually voted.

Becker sat in the White House background briefing the day before and said an official described the comparison as done with "commercial data," which without unique identifiers "is going to create a lot of false matches." CLEAR is a commercial-data product. No source says CLEAR produced the 278,000 figure, and the procurement document was published July 14, two days before the speech, so the language wasn't written in reaction to it.

The nearest thing to a live casualty list comes from a different DHS system, so treat it as an analog rather than evidence against CLEAR. Texas flagged 2,724 people as "potential noncitizens" through the revamped SAVE database; one of them, Anthony Nel, got his registration back only after renewing his passport, per NPR. A federal judge ruled the system unlawful on June 22, but a conflicting order had DHS running it again in four states by July 9.

The Bottom Line

None of this is signed. As of today there's no award record in USAspending for Thomson Reuters Special Services or West Publishing anywhere near this scale, so scope and price can still move. That cuts the other way too: a justification is what an agency writes to explain itself to its own contracting officers, before any outside pressure to soften the language, and this one names voter fraud as a reason the government needs continuous commercial access to a million people at a time.

What I'd want answered before anyone signs is what a "Voters fraud" hit actually looks like from the inside. Which fields get matched, and what threshold turns a name into a flag. DHS hasn't answered anyone who asked, Thomson Reuters says it doesn't discuss customer contracts, and Canada's NDP leader Avi Lewis is now urging Mark Carney's government to block the deal before any of it gets tested in an American court.