🚗📷 License-Plate Surveillance: Flock Safety & ALPR Cameras

Infrared cameras capture the plate, location, date and time of every vehicle 24/7 — day or night, at up to 100 mph. Independent testing (IPVM, 2021) found roughly a 1-in-10 error rate, and misreads have caused wrongful stops (see below), so vendor accuracy claims are contested.

Beyond the plate, Flock identifies vehicles by make, model, color, dents and damage, bumper stickers, roof racks, and aftermarket wheels — and its 'FreeForm' search takes natural language like 'white sports car with a racing stripe.' It can find a vehicle with no visible plate.

Flock runs a 'share-to-search' network: a department that shares its own data can search everyone else's — ~90,000+ cameras across ~7,000 networks. Over 75% of law-enforcement customers opt in; one 30-day audit logged 450,000+ nationwide searches, and single departments have queried thousands of other networks at once.

Default retention is 30 days with automatic deletion; up to a year requires approval from an elected body. Flock says 'the customer owns the data' — but its template contract grants Flock a 'worldwide, perpetual, royalty-free' license to share agency data for investigative purposes unless the department negotiates it out.

Leaked internal materials show 'Nova,' a data-fusion tool built to let officers 'jump from LPR to person' and link relatives/associates — supplementing plate data with data brokers, breach data, and public records across 20 toggleable sources. After 404 Media's report, Flock said it would drop known-breach data but kept the product.

Flock's Raven acoustic sensors flag gunshots using AI only (no human review, unlike ShotSpotter). A 2025 expansion added 'human distress' detection — the ad originally said 'screaming,' quietly amended after EFF's report. One oversight body found over 99% of gunshot alerts led to no police action.

Flock bought drone-as-first-responder company Aerodome (reported at over $300M) so an ALPR hit or a gunshot alert can auto-dispatch a drone; it later unveiled its own drone, 'Alpha.' Elk Grove PD (CA) got the first jurisdiction-wide FAA waiver to fly beyond line of sight.

The industry is now adding device tracking on top of the plate. In June 2026, 404 Media reported that Leonardo — the defense firm behind the ELSAG plate-reader line — markets 'SignalTrace,' which adds Bluetooth/Wi-Fi/RFID sensors to its readers to 'sweep up unique identifiers of mobile phones, wearables, and other Bluetooth-enabled devices,' then 'links devices that regularly travel together, correlating them to license plate.' Leonardo says it 'does not decrypt or read content.' The reporting is based on Leonardo's own product materials; no operational deployment has yet been documented — and this is an ELSAG product, NOT a Flock feature.

The device-tracking capability is real, but its reach against modern smartphones is genuinely contested. EFF's Surveillance Self-Defense notes that on current iOS and Android the Wi-Fi MAC in probe requests is randomized by default, 'which makes this kind of tracking much more difficult,' and footfall-analytics vendors report Wi-Fi capture fell to roughly 6–12% of passers-by after Apple and Google enabled it. The limit is not absolute: Bluetooth accessories, networks a phone actually joins, and cell identifiers (IMSI/IMEI) can still expose a stable ID, and randomization is 'software based… fallible.' Present the capability and the countermeasure together.

A separate, older device-identifier tool is the cell-site simulator — sold under Harris/L3Harris brands (Stingray, Hailstorm, KingFish) and by vendors like KeyW and Boeing DRT ('dirtboxes'). It masquerades as a cell tower to trick nearby phones into transmitting the IMSI (subscriber ID) and IMEI (device ID). EFF and the ACLU have documented use across the FBI, DEA, ICE, U.S. Marshals and dozens of local departments; EFF says the devices 'invade the privacy of everyone who happens to be in a given area,' capturing identifiers from up to thousands of phones at once.

Flock's own product page for 'FreeForm' lets officers find people by plain-language description of 'clothing, visible accessories' and other 'observable, non-biometric visual characteristics' (e.g. 'man in blue shirt and cowboy hat'). Flock says FreeForm 'does not use facial recognition,' but it is a person-search over video. 404 Media documented departments using it to locate people — logged queries include a target's race and clothing — and a watchdog tally (HaveIBeenFlocked) counted 6,736 FreeForm searches across 121 agencies in 2025. That sits in tension with Flock's public FAQ that its cameras 'are not designed to search for people.'

Flock's 'Condor' is a pan-tilt-zoom (PTZ) camera with AI built to detect and follow human movement, streaming live to police — distinct from the fixed rear-plate 'Falcon' units. In a June 2026 demonstration filmed by InvestigateTV, a trail-mounted Condor pivoted, followed and zoomed in on a passing reporter in real time; asked 'You're saying Flock does not track people, correct?', the Flock representative said no, while the display showed the camera tracking the reporter.

Axon's Lightpost (built with Ubicquia) clamps onto an existing streetlight and installs in under an hour — no new poles, trenching or permitting. It reads plates locally, uploads the LPR metadata to Axon Fusus for alerting and historical lookups, and lets authorized Fusus operators open a live video stream from the camera over its built-in LTE. Its sibling Axon Outpost is a fixed pole/wall camera with the same ALPR + vehicle-intelligence stack. Together they make roadside plate-and-video surveillance far cheaper to blanket across a city.

A 2025–26 wave of cities cancelled Flock over its nationwide data-sharing network, then signed Axon to run license-plate cameras instead — Denver (council voted 7–6 for 50 Axon units, with 21-day retention, no vendor-run national database, and mandatory query audit trails), Syracuse (7–1), plus Pasco County FL and Bartlesville OK. Axon's pitch is that it has no nationwide cross-agency lookup — a real difference from Flock's ~90,000-camera share-to-search pool — but the cameras still capture every passing plate and feed Axon Fusus, so the switch narrows data-sharing without ending mass plate collection.

An official at the Johnson County, TX Sheriff's Office ran two Flock searches for a woman who had a self-managed medication abortion (April–May 2025). The second reached 6,809 networks / 83,345 cameras nationwide — including states that protect abortion. The audit-log reason field read, verbatim, 'had an abortion, search for female.' The sheriff publicly called it a welfare check; court records later showed deputies had opened a 'death investigation of a non-viable fetus' and consulted the DA about charging her the same day as the search.

Local and state police ran 4,000+ Flock lookups for immigration purposes — reason fields literally reading 'ICE,' 'ICE WARRANT,' 'illegal immigration' — even though Flock has no ICE contract (records via Danville, IL). CBP separately had direct backend access to 80,000+ cameras through an undisclosed pilot; the Secret Service and Navy/NCIS also had access, and some departments did not know their data was being shared.

Illinois law (Public Act 103-0540) bars using ALPR data for immigration or abortion enforcement. The Secretary of State's audit found 46 out-of-state agencies searched Illinois data illegally, that Mount Prospect's data was queried 262 times for immigration reasons, and — in a follow-up — that Flock had granted CBP access via an undisclosed pilot in violation of state law.

Documented cases show officers using Flock to stalk romantic interests — a Sedgwick, KS police chief searched his ex 164 times and her new boyfriend 64 times (228 searches). Audits found over 20% of Austin PD's ALPR searches had no case number, and in a dataset of 11.4 million Flock searches, 14%+ of the 'reason' fields were just the word 'investigation.'

At least 60 Flock 'Condor' cameras were streaming live to the open internet with no password — exposing live feeds, 30 days of archived video, and admin controls (a reporter filmed himself on a public feed). Device flaw CVE-2025-59403 is rated 9.8 (critical). Lawmakers asked the FTC why Flock doesn't enforce multi-factor authentication after stolen police logins were found for sale.

The Institute for Justice documented at least 26 cases since 2018 of motorists wrongly detained — often at gunpoint — due to ALPR errors, most since 2023. A Flock one-letter misread in Sherwood, AR flagged a family's SUV as stolen; they were stopped at gunpoint with a ~6-week-old infant in the car. Comparable ALPR-error cases settled for $1.9M (Aurora, CO) and $495K (San Francisco).

Chicago resident Michael Williams (65) was arrested for a 2020 murder based largely on a ShotSpotter alert placing a 'gunshot' at an intersection. He spent nearly a year in jail — contracting COVID-19 twice — before Cook County prosecutors withdrew the ShotSpotter evidence and dismissed the case for lack of evidence. The MacArthur Justice Center litigated it as Williams v. City of Chicago; it became a leading example of an acoustic alert driving a wrongful arrest.

A MacArthur Justice Center study of Chicago Police data (Jul 2019–Apr 2021) found 89% of ShotSpotter deployments turned up no gun crime and 86% led to no report of any crime — 40,000+ dead-end deployments, concentrated in Black and Latino neighborhoods; Chicago's Inspector General independently found the system rarely produced evidence of a gun crime. Chicago — long the flagship city — decommissioned ShotSpotter on Sept 22, 2024 after the mayor declined to renew, part of a wave that also included Dayton, Atlanta, San Antonio, Seattle and Buffalo. SoundThinking disputes the study's methodology and cites a 97% accuracy figure it has not substantiated in a peer-reviewed study.

A BuzzFeed News review of a leaked Clearview client list found people at 2,228 agencies, companies and institutions had run nearly 500,000 face searches — including ICE (8,000+ searches from ~60 accounts) and the FBI (5,700+ across 20+ field offices), plus DOJ and CBP. Many ran through free trials individual officers activated with a government email, sometimes without their agency's knowledge or approval. ICE's largest reported purchase was a $3.75M Clearview contract.

Runs the Atlas of Surveillance and Street-Level Surveillance projects; with 404 Media, broke much of the documented oversight record. Its core critique: the danger is 'the architecture itself' — a nationwide interconnected network — 'not just how individual agencies use the technology.'

Runs a national campaign; its foundational 2013 report 'You Are Being Tracked' first called ALPRs mass location tracking. Notes fewer than 1% of scanned cars are tied to any crime, and documented the 'perpetual, worldwide' data-sharing license in Flock's contracts.

A grassroots project founded in 2024 by software engineer Will Freeman (after noticing camera clusters around Huntsville, Alabama) that crowdsources ALPR camera locations into OpenStreetMap — the dataset behind the map on this page. When Flock sent a trademark cease-and-desist in Feb 2025, EFF represented Freeman and refused it.

Litigates ALPR dragnets as Fourth Amendment violations (Norfolk; San Jose) and compiles the running tallies of wrongful stops and officer stalking.

Founded 2019 by Albert Fox Cahn; argues federal intelligence-sharing has let ICE 'sidestep sanctuary laws and co-opt local police databases and surveillance tools, including license plate readers.'

Opened Congressional investigations and urged the FTC to act. Sen. Wyden: 'I now believe that abuses of your product are not only likely but inevitable.'

A public-interest law firm (Northwestern Pritzker School of Law) that produced the leading empirical study of ShotSpotter in Chicago (89% of deployments found no gun crime) and litigates acoustic-surveillance harms, including Williams v. City of Chicago on behalf of a man wrongly jailed on a gunshot alert.

Flock Safety says its cameras help solve 'over 700,000 crimes each year' and supported 'more than 1 million criminal investigations' in 2025. Its headline 'share of crime solved' figure, though, is not a stable or comparable number: it moved from 2.5% (2022) to 10% (2024) to 20% (2026), and the base quietly shifted from 'reported crime' to 'solved/cleared cases.' Both the 700,000 and the 10% figures come from Flock-commissioned surveys, not government data — treat them as the company's own claims.

Beyond the vendor claims are specific, independently-reported cases. Boulder (CO) police credited a Flock camera with locating a 16-year-old AMBER-alert victim within minutes ('without this technology, we would send this information out statewide'); a kidnapped 1-year-old was recovered across state lines in Arizona after a Flock hit; and departments in Whiteville NC, Virginia Beach VA, El Cajon CA and Bluffton SC have credited ALPR hits with homicide arrests within hours.

In Schmidt v. City of Norfolk (Jan 27, 2026) a federal judge granted summary judgment for the city, holding Norfolk's ALPR system does not 'track the whole of a person's movements' — but expressly hedged, 'at least in Norfolk, Virginia, the answer is: not today,' warning it could become too intrusive as coverage grows; it is on appeal at the Fourth Circuit. Other rulings agree a plate 'in plain view' carries no reasonable expectation of privacy (Commonwealth v. Church, VA; Bednarz v. Illinois State Police; U.S. v. Yang, 9th Cir.). One federal ruling has since gone the other way: on Oct 1, 2026, in United States v. Kyle (N.D. Okla.), Judge Sara E. Hill held a deputy's warrantless month-long Flock/VehicleManager query was a Fourth Amendment search. Her own opinion acknowledges that 'nearly every court examining the issue to date' has held the opposite, including three Oklahoma federal courts (United States v. Acosta, N.D. Okla., Aug 2025; United States v. Brown, W.D. Okla., Aug 2025; United States v. Schoggins, E.D. Okla., Jan 2026). It binds no other court, and Flock expects it to be overturned on appeal. The split is now visible within three weeks: on Sept 10, 2026, in United States v. Riley (C.D. Cal.), Judge Virginia A. Phillips held ALPR database queries are not searches at all, on facts involving four photographs per city in an abduction investigation. A Virginia state trial court (Commonwealth v. Bell, Norfolk Circuit Court, May 2024) had earlier suppressed Flock evidence for lack of a warrant.

Flock argues ALPRs 'observe vehicles, not people,' that 'license plates are public by design,' and that its cameras create 'discrete, point-in-time observations… rather than an always-on dossier,' with 30-day default retention, hard deletion, encryption, and an audit trail of every search reason. Police groups like the IACP treat ALPR as legitimate when queries are tied to a case number, run by trained users, and audited.

A large share of the 2025–26 wave of city cancellations is openly driven by immigration/ICE concerns, not general privacy. One tracker summary found cancellations 'cluster around three themes: immigration enforcement fears, unauthorized federal access, and a collapse of trust.' Officials say so directly — a Santa Cruz councilmember tied her vote to an ICE killing and 'our Santa Cruz values,' and the EFF titled an op-ed 'San Jose Can Protect Immigrants by Ending Flock Surveillance.' (No source publishes a clean number for how many cancellations cite ICE vs. privacy — it's a documented pattern, not a tallied count.)

Non-citizens — including illegal immigrants — count in the decennial Census, which determines congressional apportionment (House seats), Electoral College shares, and the distribution of more than $2.8 trillion/yr in federal funding. Because that population concentrates in certain jurisdictions, there is a real, documented representation-and-funding stake tied to it — which is why some Republicans (the Hagerty/Cassidy bills) have moved to exclude illegal immigrants from the apportionment count.

How much this actually shifts power is disputed — and generally found small. PolitiFact rated FALSE the claim that counting non-citizens gave Democrats '24 more seats,' estimating real shifts of ~2–3 seats. Pew found that excluding unauthorized immigrants would cost California, Florida and Texas a seat each — and Florida and Texas are Republican-leaning. From 2019–2023, far more non-citizens moved to red states (~1.2 million) than blue states (~72,000). So the incentive is real, but the measured partisan effect is modest and not cleanly one-sided.

Some argue the privacy objection is selectively invoked and really tracks immigration enforcement. Pierce County (WA) Sheriff Keith Swank put it bluntly: 'Outrage over privacy and surveillance spikes the moment the tools start identifying illegal aliens who commit crimes… activists who stayed silent about other monitoring suddenly discover deep concerns for civil liberties when the targets are people in the country illegally.' The broader version ties opposition to the census/representation stake above. This is presented as an argument that is made — not as an established fact.

Opponents frame it as a universal, cross-partisan objection to warrantless mass tracking of everyone. The ACLU's Jay Stanley calls Flock 'a giant nationwide warrantless surveillance infrastructure' and stresses the data 'can be abused not just by ICE, and not just for immigration purposes' — pointing to the Texas officer who searched nationwide for a woman who'd had an abortion and to police running searches around protests. Some opponents do openly name protecting immigrants as one aim — but that is a declared policy position, which is different from a hidden partisan motive.

Being precise: the public record does NOT establish a coordinated or hidden political motive behind camera opposition as a whole. But 'not established in the public record' is NOT the same as 'false' — a hidden or mixed motive is by nature hard to document, so the absence of a paper trail is not proof that political calculation plays no role for any given person. Motives are also mixed and individual, not monolithic. The honest position is to show both the credible assertion and the documented breadth of the opposition, note that this is an emerging topic where today's picture may change, and let you weigh it.

A purely partisan read of the opposition is incomplete. The same Norfolk lawsuit is backed from both ends of the spectrum: the libertarian Institute for Justice brings it, and the libertarian Cato Institute AND the progressive ACLU/EFF each filed for the plaintiffs. Cato frames ALPRs as 'suspicionless searches that the Framers sought to prohibit,' tracing them to the 1761 writs-of-assistance fight. Red jurisdictions have dropped Flock — Sedona AZ, and the Texas towns of Bandera ('Self-governance starts at home. It's your duty to defend our constitutional rights') and Lockhart — and Republican legislators (Rep. Keith Self, R-TX) plus GOP-led Arkansas, Idaho and Montana have restricted ALPRs on Fourth-Amendment grounds.

Short answer from the record: not that anyone has documented. The confusion is understandable — anti-surveillance detectors ('Flock You' / DeFlock) locate Flock cameras by picking up the cameras' OWN emissions: Flock units broadcast a Wi-Fi SSID of the form 'Flock-[partial MAC]' plus Bluetooth beacons, which the detectors read to find them. That is people detecting the camera, not the camera capturing your device. The vendor with documented device-identifier capture is Leonardo/ELSAG's SignalTrace, which runs on ELSAG's own readers and does not integrate with Flock. So: device-MAC capture is a real, documented ALPR-industry capability — but it is not established for Flock's own cameras. 'Not documented' is not the same as 'impossible'; it is simply where the public record stands today.