Blue Light, Black Box: How Private Equity Turned Police Surveillance Into a Subscription Service

In the spring of 2023, Axon Enterprise — the company best known for manufacturing Tasers — made headlines when it proposed attaching AI-powered facial recognition to drones deployed over American cities. The backlash was swift enough that Axon's own ethics board resigned in protest, and the company temporarily shelved the idea. But the episode illuminated something that rarely gets the attention it deserves: the policing technology industry is undergoing a sweeping consolidation, driven not by public safety experts or elected officials, but by private equity firms hunting for recurring revenue in a market that never has to advertise for customers.

The clients are taxpayers. The product is us.

The Acquisition Playbook

Over the past decade, private equity has moved aggressively into what the industry calls "govtech" — government technology — with law enforcement software representing one of the most lucrative verticals. Firms including Motorola Solutions, which is backed by significant institutional investment, have absorbed companies like Vigilant Solutions (a license plate reader giant) and VideoIQ. Palantir, which went public in 2020 and counts Peter Thiel among its founders, has expanded its contracts with local police departments, ICE, and the FBI. Thomson Reuters acquired Clear, a data aggregation platform that fuses public records, social media data, and commercial databases into detailed profiles of individuals — many of whom have no idea the file exists.

The pattern is consistent: acquire a surveillance technology company, lock police departments into multi-year licensing agreements, then charge annual renewal fees that grow with each contract cycle. Because these tools are procured through administrative purchasing channels rather than public legislation, they frequently bypass the community input and democratic oversight that a new policing policy would otherwise require. A city council might debate a new use-of-force policy for months. A police chief can sign a software contract in an afternoon.

Algorithms Don't Have Precincts — But They Have Biases

The civil liberties stakes here are not theoretical. A 2019 study by the National Institute of Standards and Technology found that facial recognition algorithms misidentified Black and Asian faces at rates 10 to 100 times higher than white faces, depending on the algorithm tested. Despite this documented disparity, facial recognition is now deployed in dozens of American cities, often through contracts with private vendors whose source code is classified as proprietary — meaning neither defense attorneys nor independent researchers can audit how the system reaches its conclusions.

Predictive policing tools carry similar risks. Systems like PredPol (now rebranded as Geolitica) use historical crime data to forecast where crimes are likely to occur. But that historical data encodes decades of racially biased enforcement — more policing of Black neighborhoods generates more arrests in Black neighborhoods, which the algorithm then interprets as evidence of higher criminality, which triggers more policing. It is a feedback loop dressed up as machine learning, and it is sold to cities as objective science.

The strongest argument in favor of these tools is that they help under-resourced departments do more with less — that algorithmic assistance frees up officers for community engagement rather than patrol saturation. It is a reasonable concern, and in a world of perfectly neutral data and perfectly auditable systems, it might carry more weight. But we do not live in that world. We live in a world where the ACLU has documented wrongful arrests resulting directly from facial recognition mismatches, where Black men in Detroit and New Orleans have been jailed based on algorithm-generated misidentifications, and where the companies profiting from these errors have no legal liability for the consequences.

The Perverse Incentive Structure

Here is the structural problem that distinguishes this from ordinary government contracting: when a private equity firm acquires a surveillance technology company, its fiduciary obligation is to maximize returns for its investors. That means the business model is optimized not for accurate policing, not for community safety, not for constitutional compliance — but for expanding the scope and scale of surveillance, because broader deployment means higher licensing fees.

Consider what that incentive structure produces. A company that profits from license plate reader networks has a financial interest in those networks expanding to more intersections, more jurisdictions, more data retention windows. A company that profits from predictive policing subscriptions has a financial interest in police departments using the software as frequently as possible. There is no revenue in restraint. There is no profit in fewer scans.

This is not a metaphor. Motorola Solutions reported $8.7 billion in revenue in 2022, with its software and services segment — which includes surveillance platforms — growing faster than its hardware division. Palantir's government segment generated $903 million in 2022. These are not small players operating at the margins of law enforcement. They are among the most powerful commercial forces shaping how American policing functions.

Who Bears the Cost

The communities absorbing the consequences of algorithmic over-policing are not the communities where private equity partners live. Research consistently shows that predictive policing tools are deployed most intensively in low-income neighborhoods of color — the same communities that already experience disproportionate contact with law enforcement, that already face the highest rates of pretrial detention due to cash bail, and that already have the fewest resources to mount legal challenges when technology gets it wrong.

For a Black teenager in a city using automated license plate readers and predictive mapping, the surveillance infrastructure is ambient and inescapable. His face may be scanned walking to school. His neighborhood may be flagged as high-risk before he has ever had a single interaction with police. And if an algorithm produces a false match that leads to his arrest, the company that built the algorithm will face no civil liability, the police department will cite the technology as justification, and his family will spend money they do not have on a defense attorney to prove a machine made a mistake.

Accountability Requires Sunlight

The path forward is not technophobia. It is democratic accountability. Cities including San Francisco, Boston, and Portland have already enacted bans or strict moratoriums on government use of facial recognition — proving that local action is possible. The Facial Recognition and Biometric Technology Moratorium Act, introduced in Congress by Senator Ed Markey and Representative Pramila Jayapal, would extend those protections federally, though it has yet to advance through committee.

At minimum, every surveillance technology contract signed by a public law enforcement agency should require full algorithmic transparency, independent bias auditing, legislative approval rather than administrative procurement, and clear liability frameworks when the technology causes harm. The public pays for these systems. The public is surveilled by these systems. The public deserves to know how they work and who profits when they fail.

Private equity does not belong inside the machinery of constitutional rights — and every contract signed in the dark is a transfer of democratic power that we may not get back.