Faces in the Database: The Federal Grant Pipeline Quietly Building America's Surveillance State

The Grant That Didn't Come With a Vote

In city after city, the pattern is the same. A local police department applies for a federal grant through the Department of Homeland Security's Homeland Security Grant Program, or through the Department of Justice's Edward Byrne Memorial Justice Assistance Grant program. The application is processed bureaucratically, the funds are approved, and the department purchases facial recognition software from a vendor like Clearview AI, Amazon Rekognition, or NEC. The technology goes live. The city council never voted on it. The public never debated it. The civil liberties implications were never assessed. And by the time anyone outside the department knows the system exists, it has already run tens of thousands of searches.

This is not a hypothetical scenario. It is the documented operational reality in dozens of American jurisdictions, from mid-sized cities in the South to major metropolitan departments in states with nominally progressive governance. According to a 2022 report from the Government Accountability Office, at least twenty-one federal agencies use facial recognition technology, and federal grants have been a primary funding mechanism through which that capacity has migrated to local law enforcement. The Georgetown Law Center on Privacy and Technology has estimated that roughly half of American adults are already in a facial recognition network accessible to law enforcement.

Built-In Bias, Baked-In Harm

The civil liberties argument against facial recognition is not merely theoretical, and it is not evenly distributed across the population. A landmark 2019 study by the National Institute of Standards and Technology evaluated 189 facial recognition algorithms and found that the majority produced significantly higher error rates when identifying Black and Asian faces compared to white faces—in some cases, the disparity in false positive rates ran as high as a factor of ten. The practical consequence of that statistical disparity is not abstract. It means that a Black man is meaningfully more likely than a white man to be incorrectly flagged by an automated system, presented to a detective as a suspect, and subjected to the investigative attention that follows.

That is not a software glitch awaiting a patch. It is a structural feature of systems trained predominantly on datasets that overrepresent certain demographics and underrepresent others. And it is being deployed, right now, in communities that have spent generations fighting to be treated as full citizens by the law enforcement agencies that claim to serve them.

The wrongful arrest cases that have emerged in recent years are not outliers—they are proof of concept for what happens when biased technology interfaces with an already-imperfect criminal justice system. Robert Williams in Detroit, Michael Oliver in New Jersey, Porcha Woodruff in Detroit: all Black Americans, all arrested based at least in part on facial recognition matches that were wrong. All subjected to the full weight of the carceral state because an algorithm misfired.

The Grant Mechanism as Democratic Bypass

The federal grant pipeline deserves particular scrutiny because of how effectively it circumvents local democratic accountability. When a city council votes to purchase surveillance technology, there is a record, a debate, an opportunity for public input, and a clear line of political accountability. When a police department acquires the same technology through a federal grant administered at the departmental level, all of that accountability infrastructure simply disappears.

This is not an accident of bureaucratic design. Federal grant programs for law enforcement have historically been structured in ways that maximize departmental discretion and minimize oversight requirements. The Byrne JAG program, which distributes roughly $350 million annually to state and local law enforcement, imposes minimal reporting requirements and no specific restrictions on surveillance technology acquisition. DHS's grant programs similarly leave significant latitude to recipient agencies. The result is a shadow procurement system operating in parallel to—and largely invisible to—the democratic institutions that are supposed to govern public safety policy.

Progressive cities have begun to push back. San Francisco, Oakland, and Boston have passed outright bans on municipal use of facial recognition. Portland, Oregon enacted what advocates described as the most comprehensive facial recognition ordinance in the country in 2020, covering both government and private commercial use. But these victories are fragile. They apply to city-funded acquisitions, and the federal grant loophole remains largely intact. A department that cannot purchase facial recognition technology with city funds can, in many jurisdictions, still acquire it with federal money—and the local ordinance may offer no protection.

The Strongest Counterargument

Proponents of facial recognition in law enforcement make a case that should not be dismissed without engagement. The technology, they argue, has helped solve violent crimes, locate missing persons, and identify suspects in cases where traditional investigative methods had stalled. They point to accuracy improvements in newer systems and argue that the answer to bias is better technology, not prohibition. They contend that the alternative—slower investigations, lower clearance rates, more criminals at large—carries its own human cost.

These are real considerations. But they rest on an evidentiary foundation that is far weaker than its proponents acknowledge. Clearance rates for violent crimes in the United States have been declining for decades, and there is no robust peer-reviewed evidence that facial recognition deployment has meaningfully reversed that trend at scale. More fundamentally, the argument that a technology's benefits justify its deployment in the absence of democratic authorization, judicial oversight, or accountability frameworks is an argument that would justify virtually any surveillance tool imaginable. The question is never only whether a technology can do something useful. It is whether the community it is used against consented to its use, and whether the safeguards necessary to prevent abuse are actually in place.

They are not. In most jurisdictions using facial recognition, there are no mandatory audit requirements, no independent oversight boards, no restrictions on how long facial recognition data can be retained, and no requirement that departments disclose to defendants that facial recognition was used in their investigation.

The Democratic Stakes

Surveillance technology deployed without democratic consent is not a neutral public safety tool. It is an expression of power—specifically, the power of law enforcement institutions to expand their reach over the communities they police without those communities having any meaningful say. When that expansion is funded by federal grants that bypass local accountability structures, and when the burden of false identification falls disproportionately on communities of color, the result is a surveillance infrastructure that is neither democratically legitimate nor equitably applied.

The 2024 and 2026 election cycles will bring facial recognition policy to city councils and state legislatures across the country as public awareness grows and advocacy organizations continue to document abuses. The question of who controls the technology that watches us—and who gets to decide whether it watches us at all—is not a niche civil liberties issue. It is a foundational question about what kind of democracy we are building.

A government that surveils its citizens without their consent, funded by grants that circumvent the democratic process, and calibrated to misidentify Black faces at rates that would be considered scandalous in any other context, is not a government that can claim to serve the public it is watching.