Funded by Taxpayers, Aimed at Communities: The Unaccountable Surveillance Infrastructure Being Built in Your City Right Now

The Grant Cycle Nobody Voted On

In cities from Baltimore to Fresno, from Atlanta to Albuquerque, a quiet infrastructure buildout is underway — one that most residents have never heard of, never debated, and never approved. Federal grant programs, primarily administered through the Department of Justice's Bureau of Justice Assistance and Department of Homeland Security funding streams, have for years channeled hundreds of millions of dollars annually into local law enforcement agencies for the purchase of surveillance technology. License plate readers that log the movements of every vehicle passing a camera. Facial recognition systems that attempt to match faces in real-time video feeds against criminal databases. Predictive policing software that uses historical arrest data — data that encodes decades of racially disparate enforcement — to forecast where crime will occur and who is likely to commit it.

The procurement decisions for these systems are typically made at the departmental or mayoral level, without city council authorization and almost never with structured community input. Civil liberties organizations, including the ACLU and the Electronic Frontier Foundation, have documented dozens of cases in which major surveillance contracts were signed before the public was even aware a proposal was under consideration.

A Technology With a Documented Failure Rate

The case for these systems rests on an efficiency argument: better data means faster case resolution, smarter resource deployment, and ultimately safer streets. It is a case that deserves honest examination — and honest examination reveals it to be substantially weaker than the vendors selling these systems would prefer.

Facial recognition technology, in particular, has a well-documented accuracy disparity problem. A landmark 2019 study by the National Institute of Standards and Technology — a federal agency — tested 189 facial recognition algorithms and found that most produced significantly higher error rates for Black women compared to white men, with some algorithms producing false positive rates up to 100 times higher for darker-skinned individuals than for lighter-skinned ones. The implications for law enforcement use are not theoretical. Robert Williams, a Black man from Michigan, was wrongfully arrested in 2020 after a facial recognition system misidentified him as a shoplifting suspect. He was detained for thirty hours. The Wayne County Prosecutor's office eventually dropped the charges — but Williams had already been arrested, fingerprinted, and held in a cell. His children watched their father taken away in handcuffs from their front yard.

Williams is not an outlier. The Innocence Project and Georgetown Law's Center on Privacy and Technology have documented multiple cases of facial recognition-driven wrongful arrest, disproportionately involving Black men. Yet the companies whose algorithms produced these misidentifications — NEC Corporation, Clearview AI, Axon — face no legal liability for the consequences. The officers who acted on the faulty match face qualified immunity. The city that deployed the system faces, at most, a civil lawsuit. The vendor faces nothing.

The Predictive Policing Loop

Predictive policing software presents a related but distinct problem. Systems like PredPol (now rebranded as Geolitica) and ShotSpotter generate deployment recommendations and crime forecasts based on historical arrest data. The circularity of this methodology is its fatal flaw: if police have historically over-policed Black neighborhoods — and the data unambiguously show that they have, producing arrest records that reflect enforcement patterns as much as criminal behavior — then an algorithm trained on that data will recommend deploying more officers to those same neighborhoods, generating more arrests, which feed back into the model to confirm the original bias. The algorithm does not detect crime. It launders discriminatory policing into the language of data science.

A 2021 study published by researchers at Princeton University examined the deployment of ShotSpotter in Chicago and found that the system generated alerts that led police to overwhelmingly Black and Latino neighborhoods, with the vast majority of deployments resulting in no evidence of a shooting. The city of Chicago paid ShotSpotter more than $33 million over several years for a system that its own inspector general concluded lacked sufficient evidence of effectiveness.

The Democratic Deficit at the Heart of Surveillance Expansion

The strongest counter-argument from law enforcement and city administrations is that these tools, properly implemented and overseen, provide genuine public safety benefits — and that communities experiencing high rates of violent crime have both a need and a right to advanced investigative tools. This argument carries some moral weight. Victims of violent crime in underserved communities deserve effective law enforcement responses, and dismissing technology categorically risks overlooking tools that, under rigorous oversight, might serve legitimate purposes.

But the operative phrase is "rigorous oversight" — and it is precisely what is absent. There is no federal statutory framework governing law enforcement use of facial recognition. There is no mandatory disclosure requirement for predictive policing deployments. There is no standardized audit process, no independent review board with enforcement power, and no requirement that communities be informed before their movements are logged and their faces catalogued. The absence of these guardrails is not an oversight. It reflects the political influence of a surveillance technology industry that generated an estimated $34 billion in global revenue in 2023, according to MarketsandMarkets research, and has every incentive to ensure that the regulatory environment remains as permissive as possible.

What Accountability Requires

Several cities — San Francisco, Boston, Portland — have enacted municipal bans or moratoriums on government use of facial recognition. Oakland and Seattle have passed surveillance technology ordinances requiring council approval and public disclosure before any new system is procured. These are meaningful steps, but they are local patches on a national problem. Federal legislation establishing minimum transparency standards, mandatory impact assessments for communities before deployment, vendor liability frameworks, and an outright prohibition on facial recognition in public spaces absent individualized judicial authorization would represent the structural floor that currently does not exist.

Congressional proposals to this effect — including the Facial Recognition and Biometric Technology Moratorium Act, introduced by Representative Pramila Jayapal and Senator Edward Markey — have repeatedly stalled without reaching a floor vote. Meanwhile, the grant pipelines that fund these systems continue to flow, and the communities bearing the greatest risk of misidentification and wrongful arrest continue to have the least say in whether these systems operate in their neighborhoods.

Public safety funded by public money must be accountable to the public — and a surveillance apparatus that operates in the dark, targets the marginalized, and immunizes itself from consequence is not a safety system; it is a control system.