The DOJ Just Turned Every State Agency Into an Immigration Reporting Desk

On Tuesday, September 2, 2026, the Justice Department's Office of Legal Counsel issued an opinion that reverses a reading of federal law that stood for nearly thirty years. Under the new interpretation, any state that accepts federal money through low-income support programs, including Temporary Assistance for Needy Families and Supplemental Security Income, must report undocumented immigrants across its entire state government. Not just the welfare office that cuts the checks. The DMV. The public health clinic. The state university registrar. The unemployment insurance office. The opinion does not change a single word of statute. It changes who inside a state government is understood to be bound by it, and that shift converts routine administrative databases into a federal enforcement surface.

The legal hook is 8 U.S.C. 1373 and 1644, provisions written into the 1996 welfare reform package signed by Bill Clinton on August 22, 1996. Those sections bar state and local governments from restricting their employees from sharing immigration status information with what was then the Immigration and Naturalization Service. For decades the operating assumption, reflected in prior OLC guidance and in how state administrators built their systems, was that the reporting duty attached to the agencies administering the federal benefit. The new opinion says the condition rides on the money, and the money goes to the state, so the duty is the state's, undivided.

The Mechanics of Conditional Money

This is a story about fiscal leverage, and the leverage is enormous. TANF sends roughly $16.5 billion a year to states through a block grant that has been frozen at that nominal level since 1997, which means it has lost close to half its real value to inflation over three decades. SSI pays out around $60 billion annually, and while it is a federally administered program, more than half the states supplement it with their own payments and administer eligibility screening through state offices. Medicaid, which is the elephant standing behind both, carries a federal share north of $600 billion a year. The question of which programs count as covered "federal public benefits" under the 1996 framework is exactly the kind of question OLC opinions decide.

Every state takes this money. That is the point. A block grant is not a gift, it is a contract with a covenant attached, and the covenant can be reinterpreted by the counterparty without renegotiation. New York's TANF allocation runs about $2.4 billion. California's is roughly $3.7 billion. A state that wants to decline the terms must decline the funds, and no state legislature is going to write off billions in a budget cycle to win a jurisdictional argument. The Supreme Court drew a line on this in NFIB v. Sebelius in 2012, when seven justices held that threatening a state's entire Medicaid grant to force expansion was unconstitutional coercion, a "gun to the head" in Chief Justice Roberts's phrase. But that ruling policed the size of the threat, not the practice of attaching conditions. Conditions attached to money the state has already built its budget around are, functionally, commands.

There is also a live anticommandeering question. In Printz v. United States in 1997, the Court struck down a provision of the Brady Act that required local sheriffs to run background checks, holding that the federal government cannot conscript state officers to administer a federal program. Lower courts have split on whether 1373 survives Printz. The Second Circuit upheld it in 2020 in litigation over Byrne JAG grants. District courts in California and Pennsylvania went the other way in 2018 and 2019, holding that a statute forbidding states from directing their own employees is commandeering by another name. The new OLC opinion pushes the statute into far broader territory than the Second Circuit was reviewing, which makes fresh litigation close to certain.

Two Readings of the Same Statute

The administration's case is straightforward and not frivolous. Congress wrote in 1996 that no state may "prohibit, or in any way restrict" any government entity or official from sending immigration status information to federal authorities. The text says "any" and it says "entity." It does not say "any entity administering a federal benefit program." On a plain-text reading, the prior narrow construction was the aberration and the new opinion is a correction. Supporters will add that sanctuary policies in roughly a dozen states have functioned as a workaround for thirty years, and that a state cannot take federal money while instructing its workforce to withhold information federal law says must flow freely.

The opposing reading is equally grounded. Sections 1373 and 1644 are gag-rule provisions, not affirmative mandates. They forbid a state from blocking a willing employee who wants to share information. They do not create a duty to collect immigration status, to build a reporting pipeline, or to push data to Immigration and Customs Enforcement. Reading a prohibition on restriction as an order to report is a category change, and it happens to be the change that converts a passive statute into an active enforcement apparatus staffed by state employees at state expense. Attorneys general in California, Illinois, New Jersey, and Washington have challenged federal immigration conditions on grants repeatedly since 2017 and have won more often than they have lost. Expect complaints within weeks, and expect at least one district court injunction before the end of the year.

There is a practical objection that cuts across the legal one. State health departments have argued for years that immigration reporting requirements suppress participation in communicable disease programs and prenatal care, including among US-citizen children in mixed-status households. The Migration Policy Institute has estimated that roughly 4.4 million US-citizen children live with at least one undocumented parent. When the reporting obligation extends to every state office, the calculation for those families changes at the vaccination clinic and the school lunch line, not just the welfare office. Whatever one thinks of immigration policy, the mechanism here is a state deciding it can extract compliance by making ordinary civic transactions carry enforcement risk.

Data That Was Collected for One Purpose

The deeper story is not immigration. It is what happens to information after it is collected.

Every one of these databases was built for a narrow, defensible reason. A driver's license file exists so the state knows who is qualified to operate a vehicle. A public health registry exists to track outbreaks. A university enrollment record exists so the registrar can issue transcripts. Each of those systems was sold to the public on its stated purpose, and in each case citizens supplied information because the purpose seemed reasonable and the alternative was going without a driver's license or a diploma.

Tuesday's opinion demonstrates the flaw in that bargain. The purpose limitation was never load-bearing. It was an administrative convention, and administrative conventions are revised by memo. No state legislature voted on this. No court ordered it. A legal opinion from an executive branch office reclassified the permissible uses of state-held data, and by force of the funding condition, every state must now conform. Twenty-eight states issue driver's licenses or equivalent cards without regard to immigration status. Those programs were enacted specifically on the promise that the data would not be used for immigration enforcement. That promise was made in good faith by people who did not control the future interpretation of a 1996 funding condition.

This is the pattern worth naming, because it is not confined to immigration. Data collected under one justification migrates to another when the political cost of the migration falls below the political benefit. The Social Security number was created in 1936 with printed assurances that it was "not for identification." Currency transaction reports under the Bank Secrecy Act of 1970 began at a $10,000 threshold that has never been indexed to inflation, so that same report now triggers at what is roughly $850 in 1970 dollars, a fivefold real-terms expansion accomplished by doing nothing at all. Suspicious activity reports filed with FinCEN have grown from roughly 62,000 in 1996 to more than 4.6 million a year. The infrastructure never shrinks. The threshold for its use only falls.

The Monetary Parallel

Here is where this connects to money, and I will not hedge on it.

A financial system in which every transaction is legible to the state is a system in which the terms of access can be revised by memo. That is the same structure as the DOJ opinion, applied to the ability to buy food and pay rent. The current banking system already runs on this logic. Know-your-customer files, transaction monitoring, and the correspondent banking network mean a person's economic life is a queryable record held by intermediaries who are legally obligated to answer when asked and legally barred from telling the customer they were asked. Operation Choke Point between 2013 and 2017 showed that this pipe can be used to squeeze legal-but-disfavored businesses without any legislation whatsoever. Canada's invocation of the Emergencies Act in February 2022 froze roughly 280 accounts in a matter of days on the strength of an executive determination.

A central bank digital currency would make this native rather than incidental. The Bank for International Settlements counted 134 countries exploring CBDCs as of 2025. Programmable money with identity attached is the DOJ opinion generalized: a system where the permissible uses of your financial data, and eventually of your money itself, are set by whoever currently holds the pen. The defense against that is not better privacy law, because the law is exactly what got revised on Tuesday. The defense is money whose rules are not revisable by opinion. Bitcoin's supply schedule cannot be reinterpreted by an Office of Legal Counsel. A self-custodied key does not have a compliance department that can be served with a demand the holder never learns about. This is not a claim that Bitcoin solves immigration policy or that everyone should hold their savings in it. It is a narrower and firmer claim: the class of systems whose rules change when the political wind shifts is a different class from the one Bitcoin belongs to, and the events of September 2 are a clean demonstration of why that distinction matters to anyone, of any political persuasion, who might someday be on the wrong side of a reinterpretation.

The Austrian point underneath is that discretionary authority tends to expand toward the limit of what it can accomplish, because the cost of expansion is borne by whoever is targeted and the benefit accrues to whoever is expanding. Rules that bind the rule-maker are rare and fragile. Rules enforced by physics and distributed consensus are neither.

What to Watch

Litigation before November. Expect multistate complaints from California, Illinois, New York, New Jersey, and Washington, filed in the Ninth and Second Circuits, arguing anticommandeering under Printz and Spending Clause limits under NFIB v. Sebelius. A preliminary injunction covering at least the plaintiff states is more likely than not before the end of 2026. A circuit split is close to certain given the Second Circuit's 2020 ruling, which puts this on a track toward Supreme Court review in the 2027 or 2028 term.

Which agencies actually comply. The opinion binds states, but compliance is administered by thousands of separate offices with separate IT systems. Watch whether state health departments and public universities are treated as covered entities in the first round of guidance from HHS and the Social Security Administration. That guidance, not the opinion itself, determines the practical scope.

Funding threats as the enforcement tool. The opinion has no teeth without a withholding action. Watch for the first formal notice that a state's TANF or SSI administrative funding is at risk. The dollar figure attached to that first notice will signal whether the administration is testing the coercion line the Court drew in 2012 or staying well inside it.

Data collection changes upstream. The most consequential response may be states declining to collect immigration status at all, on the theory that information never gathered cannot be reported. If several states move that way, expect a follow-on federal effort to mandate collection, which raises the commandeering problem in its starkest form.

Second-order effects on financial surveillance. Watch whether the same "the condition rides on the money" logic appears in Treasury or FinCEN guidance. The reasoning is portable. Any institution taking federal support, including deposit insurance and Federal Reserve account access, is exposed to the same argument. If that argument surfaces in a banking context within the next year, the ceiling on where this goes is considerably higher than immigration policy.

---

Source: Guardian US