A detailed view of hands writing on a form placed on a clipboard, with a striped fabric partially visible in the background.
getty
Many families applying for cash assistance assume the information they submit is used only to determine eligibility. A federal policy change suggests that assumption now deserves a closer look.
In June, the Administration for Children and Families updated the Temporary Assistance for Needy Families data system. Beginning September 30, the change will allow Social Security numbers, home addresses, immigration status, and other information families provide when applying for cash assistance to be shared with Department of Homeland Security systems — without a warrant, a court order, or the caseworker knowing. More than 20 state attorneys general are now challenging the policy in court. The legal challenge raises new questions about how government technology governs the lives of people who rely on public benefits.
The notice was published as a routine Privacy Act notice. It allows the routine uses of welfare applicant data containing Social Security numbers, home addresses, and immigration status to be shared with the Department of Homeland Security for uses that go beyond administering benefits. The policy does not require warrants, court orders, or individualized suspicion before sharing. It also doesn’t establish a process for recipients to learn they’ve been flagged, correct an erroneous match, or contest an adverse action.
By Aug. 3, more than 20 states had joined legal challenges seeking to block the change. Attorney General Letitia James of New York was among the first to file suit. California Attorney General Rob Bonta, along with Attorney General Nick Brown of Washington state, joined the case, as did officials from Kentucky and Pennsylvania. The states argued that this change violates the Administrative Procedure Act, the Computer Matching and Privacy Protection Act of 1988, and the Constitution’s spending clause.
Bonta announced the suit and said, “The Trump administration is exploiting a program designed to ensure children do not go hungry and to help needy families get back on their feet in order to fuel its mass surveillance effort. It’s cruel, unnecessary, and illegal.”
How Benefits Became Data
The story didn’t start in June 2026. It began in the late 1990s when the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 required states to transition to electronic benefits transfer cards. The delivery of benefits itself would now generate digital records.
The same law also required states to set up automated matching across agencies. Congress passed the Computer Matching and Privacy Protection Act of 1988 to limit this kind of sharing among agencies. The law required agencies to publish matching agreements and get approval from a data integrity board. PRWORA changed that framework. Matching became a condition for receiving federal funds. The legal architecture that was supposed to limit data sharing became the price of admission to the welfare system.
What followed was incremental expansion. EBT cards replaced food stamps. Transactions became data, and state agencies began sharing that data with private contractors to detect fraud. Contractors ended up storing behavioral information for millions of low-income Americans. By the mid-2000s, someone applying for food assistance could create a record visible to multiple agencies governed by overlapping matching agreements that very few people outside of the bureaucracy read.
Take Indiana as an example. In 2006, the Indiana Family and Social Services Administration outsourced welfare processing to IBM. Under the new system, application errors, missing signatures, or minor mistakes were categorized as “failure[s] to cooperate.” In her 2018 book, Automating Inequality, Virginia Eubanks documented that roughly 1 million applications for SNAP, cash assistance, and Medicaid were denied over three years. Many of those denials affected families who were actually eligible. The system followed its programming with very little individual human review of the applications.
When Program Integrity Collides With Privacy
Governments have a legitimate interest in protecting public programs, but fraud prevention is not the same as a license to repurpose benefit data. Identity theft, duplicate payments and administrative errors cost taxpayers money and undermine confidence in programs on which millions of people rely. Proposed safeguards shouldn’t make catching actual fraud harder. A 2022 Center for Law and Social Policy analysis by Parker Gilkesson, a former SNAP caseworker, found that intentional SNAP fraud is rare. It accounts for 0.1% of SNAP issuances, and the total overpayment rate is 0.9%. Infrastructure designed to prevent fraud has grown much larger than originally intended.
But accurate payments do not require treating enrollment in a public benefit program as blanket consent for unrelated enforcement. Nor does program integrity, by itself, answer whether information collected to administer benefits should later be available for unrelated immigration enforcement. Algorithmic flags also require meaningful review, particularly when they can trigger investigations or adverse actions. The harder question is where program integrity ends and surveillance begins. As the amount of data available to the government expands, that boundary becomes increasingly important.
The Bigger Digital Infrastructure Problem
This isn’t only a welfare story. Across government, agencies increasingly rely on interconnected data systems for program administration, fraud detection, and enforcement coordination. The issue raised by the TANF lawsuit extends beyond one benefits program which is when governments build digital infrastructure, what limitations should apply to how that infrastructure can be used?
The Computer Matching Act created procedural protections for matching records across agencies, including requirements that agencies publish matching agreements and obtain approval from data integrity boards. Those protections were designed around discrete exchanges of information between agencies. Today, government data infrastructure also includes proprietary systems operated by private vendors and increasingly interconnected databases. Oversight built around individual matching agreements may not answer how information moves through these systems or how that infrastructure should be governed.
The Choice Governments Face Now
None of this was inevitable, nor was it designed at any single moment by anyone. It is the cumulative effect of decisions made over decades, each of which had its own defensible justification. The June notice did not create surveillance on welfare recipients — it expanded the possibilities of surveillance. Whether that expansion stands will be determined by the lawsuit.
Public infrastructure is rarely neutral. It encodes choices about efficiency, privacy, accountability and power. Governments have already built the digital systems that support public services, so the debate now centers on which values those systems will uphold and who gets to decide.
Leave a comment