Contents

1. Where Apex stands todayThe verified facts of the current contract, the June 2026 federal access disclosure, and the town’s crime profile
2. 1. Where Flock came fromFounded 2017 to sell cheap cameras to neighborhood associations, and the consent model that was set then
3. 2. From HOA amenity to national networkHow a neighborhood camera became a 120,000 camera searchable network, and what Apex actually bought
4. 3. The case for the camerasNamed rescues and arrests, the vendor’s aggregate claims, the independent research, and the percentage question answered honestly
5. 4. The case against, in names and numbersWrongful stops at gunpoint with settlement figures, the package theft case, accuracy at the alert level, insider misuse, federal access, security, and scope creep
6. 5. The abuse record, countedThe full Institute for Justice abuse database tabulated by category, year, state, vendor and outcome, with the five North Carolina records
7. 6. What the Town’s own records showThe Police Department’s own success file, the camera count nobody could confirm, the ten camera addendum, an unresolved security report, and the April federal access requests
8. 7. The legal theory, and why nobody votedThe Fourth Amendment arguments on both sides, North Carolina’s statute, the absence of any consent requirement, and how vendors structure contracts to avoid a vote
9. 8. What this could cost ApexSettlements actually paid, why the Town and not the officer or the vendor holds the risk, the contract’s liability cap, insurance gaps, and the 2024 utility billing attack
10. 9. What the other cameras teachRed light cameras in the United States and North Carolina, and how European law treats the same technology
11. 10. Why this is not like your phoneConsent, aggregation, legal process, and what deletion actually means
12. 11. Axon, the proposed alternativeWhat Axon actually sells including police body cameras and drones, the consolidation and vendor lock-in record, who owns the company, and the Trump investment claim checked against the filings
13. 12. Flock versus Axon, term by termThe published contract terms side by side, Flock’s August 2026 changes, and why the swap may not solve the problem that started this
14. 13. Four options, and what each requiresIncluding the fourth option the three option framing leaves out
15. 14. Executive summaryThe whole briefing in one page
16. 15. The national argumentWhy governors, sheriffs, senators and police chiefs in both parties are pushing back in 2026
17. 16. A field guide for Apex residentsVerified contacts for the Mayor, Council and Police Chief, why the Mayor is the key recipient, and three ready-to-send emails on a two week cadence
18. Appendix AQuestions for the vendor, the Chief and the Town Attorney
19. Appendix BFive claims to stop using, and what to say instead
20. Appendix COpen questions and the document that settles each
21. Appendix DSourcing note and method
22. Appendix EFull source list, with links to everything cited

Where Apex stands today

Everything in this section is drawn from Town records as reported by The Peak Weekly, WRAL, Hoodline and Indy Week. Two items need to be confirmed against the Town’s own files before this briefing is quoted publicly, and they are marked.

The contract

The consent agenda fact is the single most important item on that table. A consent agenda is where a council places items it does not expect to debate. There was a vote, and it was unanimous, but there was no stand-alone discussion and no public hearing. A follow up work session in February 2025 addressed policy safeguards, which is to the Town’s credit and is also the opposite of the usual order. This renewal is the first genuine public deliberation the Town has held on the technology, which is an argument for taking the time rather than for hurrying.

The federal access episode

• On June 9, 2026, at a meeting attended by more than 150 residents, the Town disclosed that two federal agencies had been mistakenly enabled on its Flock account: the U.S. Postal Inspection Service and a U.S. Air Force base police department. Access has since been terminated.
• The Town reported three nationwide searches by the Postal Inspection Service, tied to an embezzlement case.
• DeFlock Apex, an advocacy group, reports a far larger figure: 9,683 searches by Postal Inspection Service accounts reaching Apex cameras between January and April 2026, plus 17 from the Air Force base. The gap is almost certainly definitional, searches initiated by that agency versus searches whose scope reached Apex cameras. The Council should require the Town to reconcile the two numbers in writing, with a definition of each term. Until it does, neither figure should be used in public.
• This matters because of what was said earlier. In April 2026 Apex police leadership told Indy Week: “We do not share with federal entities, hard stop… we don’t share our data with the FBI, we don’t share our data with DHS, we don’t share our data with ICE.” The June disclosure contradicts that statement. The most likely explanation is that nobody at the Town knew, which is the more serious problem of the two.

What Apex asks the system to do

• 2024 reported crime, from FBI derived data, population 75,977: 38 violent offenses and 647 property offenses. Property crime is 94.5 percent of all reported crime in Apex.
• Apex’s violent crime rate is roughly seven and a half times below the North Carolina rate; its property crime rate is roughly 2.3 times below it.
• The category breakdown inside that property crime number is the most decision relevant gap in this briefing. The strongest research supporting plate readers finds an effect on motor vehicle theft specifically. If Apex’s property crime is mostly larceny from vehicles and retail theft along NC 55 rather than auto theft, the best available evidence for the technology does not reach the town’s actual crime profile. The Police Department’s annual report or an SBI agency level query would settle this in a day.

The calendar

Mayor Gilbert has framed the choice in three parts: “Do we continue with ALPR flock in our town? Do we consider moving to a different provider for ALPR? Or do we just discontinue ALPRs in Apex?” Section 13 argues there is a fourth.

1Where Flock came from

Flock Safety was not built for police departments. It was built to sell cameras to homeowners associations, and the way it handles consent was settled in that first market and has never really changed.

The founding

• Founded in 2017 in the Atlanta area by three Georgia Tech alumni: Garrett Langley (electrical engineering, now CEO), Matt Feury (computer science, CTO) and Paige Todd (management, Chief People Officer). The company went through Y Combinator.
• The origin story as the company tells it: car break-ins in Langley’s own neighborhood. Existing license plate recognition systems cost around $25,000, which he considered impossible for a neighborhood to fund. He built a cheaper camera. Todd cold contacted neighborhoods with the same complaint, two bought on the spot, and the first customers wired money to Langley’s personal account.
• Nothing in the independent reporting contradicts that origin. What independent reporting does complicate is the company’s later evidence and public statements, which Sections 3 and 4 cover.

The first market was HOAs, and it stayed that way for years

• Homeowners associations were more than 40 percent of Flock’s business as late as around 2021, before police contracts overtook them (Fortune, August 2026). Selling to neighborhoods rather than only to police was the deliberate difference from competitors.
• What HOAs actually used the cameras for, from contemporaneous reporting and Flock’s own HOA marketing page: home and car break-ins, package theft (Flock’s phrase is “detect and deter porch pirates”), illegal dumping (one California HOA used them to stop mattresses and construction debris being dumped on a roadside), street racing, and monitoring the entrances and exits of the subdivision. One Fort Worth HOA used footage for a dog attack and for property damage claims.
• They were also sold as a cheaper substitute for physical security. Flock materials cited roughly $40,000 a year saved versus installing a gate. A Brentwood, California association said it had been spending $300,000 a year on security patrols and eliminated a patrol shift after installing cameras.
• Early pricing was $1,500 to $2,400 per camera per year, plus a one time installation charge of a few hundred dollars, with maintenance, hosting and cellular service included.

The consent model, which is the part that matters

• HOA cameras were aimed at public roads at the boundaries of the neighborhood. They photographed everyone who drove past, resident or not.
• Flock’s position then and now is that the customer controls the data: “Your HOA owns all footage and data,” and law enforcement gets access only if the association opts in. The association’s board makes that decision, not the residents.
• Residents could ask to be added to an opt out or “safe list.” That is not an opt out from being photographed. The camera still reads the plate and then deletes the record. Non-residents driving the same public street had no mechanism at all.
• Sharing could be extensive. Oakland’s Lakeshore Homes Association shared its HOA funded camera data with nine outside agencies, including police in six other cities, the California Highway Patrol and the San Francisco District Attorney. It could not share with Oakland’s own police department, because Oakland had no Flock agreement.

Two documented HOA episodes worth knowing

• Oakland, February and March 2024. In roughly one month, the Lakeshore association’s cameras scanned 347,000 plates, produced 259 hotlist matches, and were the subject of 5,540 law enforcement searches, of which the California Highway Patrol alone ran 3,476. Oakland’s Privacy Advisory Commission said the installations happened with “zero transparency.”
• North Fort Worth, 2024. Flock installed dozens of cameras for HOAs without required city permits, and reporting found it had also skirted right of way rules and state private security licensing. The city issued a cease and desist. An assistant city manager’s question is the one a town asks eventually: “We need to find out if Flock is willing to assume liability for those cameras.”

Why this history is directly relevant to Apex

• Private deployments in and around Apex, reportedly including big box retailers and WakeMed, sit outside the Town’s contract but inside the same searchable network. Cancelling the Town’s contract does not remove them.
• The Town’s regulatory levers over private cameras are limited and legally untested. Chapter 74D, the Security Systems Licensing Act, is a state licensing scheme that plainly covers camera installation, and installer licensing is state occupied. In October 2023 a Wake County Superior Court judge ordered Flock to stop installing systems in North Carolina unless it used a licensed third party, finding the unlicensed work “an immediate and direct threat to the public health, safety and welfare.” Flock obtained a license in July 2024.
• A more practical lever is right of way. Private cameras in the public right of way need an encroachment agreement, from NCDOT on state maintained roads and from the Town on town maintained streets. Siting control is often available where regulatory authority is doubtful.

2From HOA amenity to national network

The product Apex bought in January 2025 is not the product Flock sold in 2018. The camera is close to the same. What changed is that it is now a node on a searchable national network, and the network is the thing the renewal decision is actually about.

The expansion, by date

Scale today

• Roughly 120,000 cameras across 49 states and 5,000 to 6,000 communities, serving about 4,800 law enforcement agencies and about 1,000 businesses. Camera counts from independent trackers ranged from 80,000 in February 2026 to 120,000 in August 2026, so treat any single figure as approximate.
• About 20 billion vehicle detections per month. This figure is company reported and has never been independently audited. At 120,000 cameras it implies roughly 5,500 reads per camera per day, which is high. Use it, but attribute it to the company.
• Revenue crossed $500 million in annual recurring revenue in the first half of 2026. Valuation is roughly $8.3 to $8.4 billion as of April 2026.
• Named investors: Andreessen Horowitz (led the 2021 and 2025 rounds), Tiger Global (led the 2022 round at a $3.5 billion valuation), Bedrock, Meritech, Matrix Partners, Founders Fund, Y Combinator, and, notably, Axon, which participated in the Series C. Section 11 returns to that last point.
• Flock leases. It owns and operates every device. The Town is not buying cameras, it is subscribing to a service.

The network, and why enrolling is not a bilateral act

• Flock’s National Lookup Tool lets an agency in one state search data collected in another, on the condition that it shares its own data in return. The ACLU’s summary of the incentive is blunt: “You show me yours, I’ll show you mine.”
• About 75 percent of Flock’s law enforcement customers are enrolled. The ACLU’s Massachusetts analysis found that a participating department’s data becomes reachable by roughly 7,000 agencies and organizations nationwide.
• The volume is not theoretical. One 30 day audit window in spring 2025 recorded more than 450,000 searches of the nationwide database. In Virginia, cameras were searched 7 million times in twelve months.
• Reported figures for Apex specifically, from DeFlock Apex and therefore unverified, describe 994 to 1,070 outside organizations with access as of spring 2026, and single searches reaching more than 6,800 networks and 100,000 cameras. The Council can settle this precisely by requesting the current network sharing roster and one month of audit logs.

The honest version of both cases

• For the network: a stolen car crosses a town line in minutes and a county line in twenty. A hotlist is close to useless if it stops at the town limit. Multi jurisdiction reach is what produced almost every rescue in Section 3, including a Utah abduction solved in Colorado and a Tennessee case solved in North Carolina.
• Against the network: to make those hits possible, everything is retained about everyone, and the sharing default converts a town’s cameras into an investigative tool for thousands of agencies whose policies, politics and oversight the town does not control and cannot inspect.
• Both statements are true at the same time. The question in front of the Council is not which one to believe. It is which configuration of retention, sharing and audit the town is willing to defend.

3The case for the cameras

This section is deliberately the strongest available version of the argument for renewal, because a briefing that understates it is not useful to anyone. The cases are real, verifiable and in several instances almost certainly saved a life.

Documented cases, with city, date and outcome

• Boulder, Colorado, September 2025. A 16 year old girl was forced into a pickup. A Flock camera in Thornton logged the suspect vehicle shortly afterward and she was found safe at a Thornton store the next morning. A 19 year old man was charged. Chief Stephen Redfearn: “Within minutes, having an idea of where that vehicle was recently is so important.”
• Aurora, Colorado, December 14, 2025. A 14 month old child in a stolen vehicle was recovered. Flock says it happened in under ten minutes and before an Amber Alert could be issued; local television independently confirms the rescue and that police credited the readers. Aurora police: “Without this technology, this outcome could have looked very different.”
• Vance County to Charlotte, North Carolina, July 2026. A child was reported missing from Henderson. Vance County deputies and Charlotte Mecklenburg police used Flock to track the suspect vehicle across counties. A 44 year old Wake County woman was arrested in Charlotte and held without bond; the child was placed in protective custody. This is the closest case geographically to Apex.
• Lenoir City, Tennessee to Jackson County, North Carolina, August 2026. A missing 13 year old, groomed online, was located roughly 300 miles from home within about four hours of police receiving video, using plate readers to identify the vehicle. Two men were charged. The chief called the technology “absolutely a game changer in policing.”
• Copperas Cove to Killeen, Texas, July 2026. After a 29 year old woman was shot and killed, a partial plate and vehicle description were entered into Flock. The system alerted automatically when the vehicle passed an intersection and a suspect was arrested after a brief pursuit.
• Indian River County, Florida to West Virginia, August 2026. A murder suspect was tracked across state lines by plate readers and jailed. Sheriff Eric Flowers: “So for everybody who wants to talk about how bad this system is, we got a murderer behind bars because of this.”
• Petaluma to Humboldt County, California, August 2026. A missing at risk adult with memory problems was tracked north through four counties in real time and found safe the same night, more than 150 miles from home.
• Roanoke, Virginia, April 2026. A missing 75 year old man with dementia was located in his car, out of gas and lost, and escorted home.
• El Cajon to Chula Vista, California, November 2025. During a burglary with the occupants at home, the suspects stole the victims’ SUV. Dispatchers tracked it live and two people were arrested in about three hours; the vehicle was returned.
• Albuquerque, New Mexico, 2026. The police department credits plate readers, alongside other tools and detective work, in 14 homicide clearances this year. Note the detail that cuts both ways: Albuquerque says it does not use Flock and does not share its data, which suggests the capability, not the network, is doing the work.

North Carolina program numbers

• Raleigh: the first 25 Flock cameras, installed June 2022, produced 116 alerts for wanted people and 41 arrests in six months. Chief Estella Patterson said they “have been instrumental in, when a crime happens, being able to locate the vehicle of the offender.”
• Garner: 15 cameras produced 144 alerts, 167 charges and 43 recovered vehicles.
• These are the numbers a neighboring department will put in front of this Council, and they are legitimate. What they measure is activity and arrests, not crime reduction, and they carry no denominator.

What Flock itself claims, and what the claims mean

• 2024: “Flock Safety technology is instrumental in solving 10 percent of reported crimes in the U.S.,” “700K crimes each year are solved using Flock,” and “one additional Flock camera per sworn officer correlates with a 9.1 percent increase in clearance rate.” Basis: 123 agencies, April to June 2023.
• 2025 Impact Census: approximately one in five resolved cases in Flock jurisdictions involved Flock technology; approximately one million cases supported; approximately 10,000 missing persons per year located, with a stated range of 8,000 to 19,600.
• Flock’s own caveat, quoted: these figures are “intended to show order-of-magnitude impact and should be read as directional estimates rather than audited totals.” The data is self reported survey data, and agencies without tracking systems supplied percentage estimates rather than counts.
• The denominator quietly changed between the two claims. Ten percent of all reported crime in the United States is not the same measurement as one in five cleared cases in towns that deploy Flock. They are not comparable, and the second is not an improvement on the first.
• The criminologist named as validating the 2024 study, Professor Johnny Nhan of Texas Christian University, later told reporters that police data is “too varied and incomplete for us to do any type of meaningful statistical analysis on them,” and that “I personally would have done things much differently.” He has moved away from further Flock research.
• Records reviewed by Techdirt show Flock recommended which departments the researchers should contact, steering toward small towns with already low crime rates where any reduction looks dramatic in percentage terms, and that researchers were asked for data showing a “big swing.”
• Forbes checked two specific city claims in February 2024. In San Marino, California, Flock claimed a 70 percent reduction in all crimes and an 80 percent drop in residential burglaries; residential burglaries had actually risen 5 percent and Part 1 crimes were flat. The city’s police chief said the claim “isn’t accurate” and added, “I definitely need to talk to their marketing folks.” In Fort Worth, Flock claimed a 22 percent violent crime decrease in one neighborhood while comparable crimes rose 5 percent citywide.

What independent research finds

• Randomized trials find no deterrent effect. Lum and colleagues (2011, Journal of Experimental Criminology) randomized 30 hot spots across Alexandria City and Fairfax County, Virginia. No statistically significant difference in crime during or after plate reader patrols, and no offense specific effect on auto crime. An earlier randomized trial (Taylor, Koper and Woods, 2010) found more hits and more recoveries but no reduction in vehicle theft versus control.
• Clearance effects are directionally positive and statistically weak. Koper and Lum (2019, Police Quarterly) studied roughly 100 fixed readers in one large city. Auto theft and robbery clearances improved where readers were concentrated, but “these changes were not statistically significant in multivariate analyses.”
• The strongest study supporting the technology, stated fairly. Adams and Mourtgos (2026) compared 216 agencies that deployed fixed readers at different times against roughly 1,800 control agencies, using FBI crime data. Motor vehicle theft fell about 11 percent in the first year, motor vehicle theft arrests rose about 16 percent, and stolen vehicles were recovered about six hours sooner. The caveats matter: it is a working paper that has not been peer reviewed, the authors note arrests were already trending upward before the cameras went live, overall recovery rates did not substantially improve, and the findings are confined to vehicle theft. It is not evidence about violent crime or overall clearance.
• The largest deployment in the country has flat clearance rates. Atlanta hosts more than 5,000 Flock cameras inside a surveillance program whose integrated camera count grew roughly eightfold. Using FBI incident data from 2021 to 2025, clearance rates were flat or falling across every major offense category, with homicide clearance dropping from 53.4 percent to 48.0 percent. Caveat, stated plainly: this is a single city before and after comparison with no control group, so it cannot rule out staffing and caseload effects. It is still real federal data from the biggest deployment in America, and it is inconsistent with the marketing.
• The George Mason University evidence review summarizes the pattern in four words: efficiency does not equal effectiveness. Readers scan plates faster and more even handedly than officers can, and that operational gain has not translated into measurable crime prevention.

The percentage question, answered honestly

The request behind this briefing asked for exact counts of ALPR assisted cases against crime statistics, and what percentage of crime is solved or impacted. That number does not exist, and it is worth being precise about why.

• National clearance baseline, FBI 2024: 43.8 percent of reported violent crimes and 15.9 percent of reported property crimes were cleared.

• No rigorous national figure for ALPR contribution exists, and it cannot be derived from public data. The FBI’s incident based reporting system has no field for investigative method. No jurisdiction records ALPR contribution as a case attribute. Seattle’s Inspector General found it could not even determine what records a search returned, let alone whether they mattered: “Without a log of responsive records returned per search, it is not possible to determine the effectiveness.”
• The Virginia State Crime Commission, a state body, reached the same conclusion in its 2024 report on law enforcement use of ALPR: “Most of the information on ALPR successes, errors, and misuses is anecdotal.”
• The one publicly audited chain from scan to outcome is Austin, Texas, audited by the City Auditor for March through December 2024: 75,000,000 plates scanned, 13,122 alerts, 165 arrests, and roughly 134 prosecutions. That is about one arrest per 455,000 plate reads. Both readings of that number are fair. The yield per scan is extraordinarily thin, and 165 arrests plus 134 prosecutions in nine months is not nothing.
• The right ask for Apex is the Austin chain applied locally, published on a fixed schedule: plates scanned, alerts generated, stops made, stops released because of a data error, arrests, prosecutions, convictions. That is the only way this Council obtains numbers about Apex rather than numbers about a vendor’s marketing.

4The case against, in names and numbers

The harms are not hypothetical and they are not evenly distributed. Two thirds of the documented wrongful stops involved officers with guns drawn, and the best documented harm category is not misreads at all. It is officers using the system to track women.

Wrongful stops at gunpoint, with what was paid

• Brittney Gilliam, Aurora, Colorado, August 2, 2020. $1.9 million settlement. The reader matched her plate number to a stolen motorcycle bearing Montana plates. She was driving an SUV with Colorado plates. Officers acted on the number without verifying the state or the vehicle type. Gilliam and four children, including her six year old daughter, were ordered out at gunpoint and made to lie face down on hot pavement; the children were handcuffed. The plaintiffs’ attorney’s line is the one to remember: “A robo-cop mentality will lead to huge liability.”
• Denise Green, San Francisco, March 2009. $495,000 settlement. The reader misread a 3 as a 7. Her burgundy Lexus was flagged as a stolen gray GMC truck, a vehicle type mismatch nobody checked. Four to six officers pointed guns at her, one with a shotgun. The Ninth Circuit revived her suit by framing the question exactly right: whether it was reasonable to stop her “without making an independent visual verification of the license plate.”
• Brian Hofer, Contra Costa County, California, Thanksgiving Day 2019. $49,500 settlement. Not a misread. His rental car had been stolen in October and already recovered, but nobody cleared it from the hotlist. Three deputies, guns drawn. Hofer identified the compounding factor: deputies are trained that “every single stolen vehicle is a high risk stop, and that guns are necessary.” No policy changes were required as part of the settlement.
• Morristown, Tennessee, June 2024. A Flock camera read the letter O as the numeral 0. Two grandparents were held at gunpoint while a three year old watched from the car.
• Sherwood, Arkansas, February 2026. A one character misread. A couple was detained at gunpoint with a six week old baby left alone in the car.
• San Diego, November 2025. Officers searched Flock by make and model rather than by plate, for a red Alfa Romeo. The system matched an identical car five miles away. Two men were arrested at a cigar lounge and one was jailed for nearly a month before charges were dropped. They are now suing the City for false arrest, wrongful detention and negligence, arguing the plate number and apartment security footage would have cleared them immediately.
• Volusia County, Florida, April 2026. A driver was jailed 13 days after a Flock capture was mistaken for a vehicle in a fatal accident.
• North Carolina has two documented cases. Charlotte, July 2022: officers entered incorrect data and a woman was arrested. Fayetteville, 2023: officers misinterpreted plate reader data, guns were drawn, and the case settled.

The size and shape of the error record

• The Institute for Justice documents 31 incidents since 2018 of innocent motorists pulled over, detained at gunpoint or jailed because of stationary plate reader errors, and calls it “most likely an undercount.” In nearly two thirds, officers drew guns before realizing the error. The longest wrongful jailings in that record are 28 days and 13 days.
• Source caveat, stated up front: the Institute for Justice is a libertarian public interest law firm currently litigating against Flock. Its index is advocacy maintained, not a neutral audit. Individual entries are sourced to news reports and public records, and the ones checked for this briefing held up.
• The cause split is the most important finding in the whole error record. By the Institute’s own coding, roughly one third is machine error and two thirds is human error: the wrong plate typed into a hotlist, a recovered vehicle never cleared, an alert misinterpreted. A more accurate camera does not fix most of this. Only verification requirements, hotlist hygiene and audits do.
• The recurring failure modes are worth naming, because each has a specific policy answer: character confusion (0 and O, 2 and 7, 3 and 7, 8 and 9); the right number on the wrong state’s plate; a plate match with a vehicle type mismatch; a stale hotlist; a bad manual entry that proves nearly impossible to remove, which left several Colorado drivers repeatedly pulled over; temporary tag handling failures; and non-plate “signature” matching that finds the wrong car of the same make and color.

The package theft case, in full

• Chrisanna Elser, Bow Mar and Columbine Valley, Colorado, reported March 2026. She was accused of stealing a $25 package from a porch. Flock cameras had logged a forest green Rivian pickup matching hers passing through between 11:52 a.m. and 12:09 p.m. on the day of the theft. On that basis a sergeant issued a court summons.
• Police came to her home. She tried to show exculpatory evidence and the officer dismissed it. She exonerated herself by assembling Google location history placing her at a tailor’s appointment more than a quarter mile away, her vehicle’s own camera and GPS records, and other Flock footage showing her truck parked elsewhere.
• The chief voided the summons before it entered the court system. The officer received a formal reprimand and additional training.
• The structural point is the one to make at a council meeting. A camera hit arrived carrying enough apparent authority to generate a criminal summons, and the entire burden of disproof fell on an innocent person who happened to own a dashcam, a GPS equipped truck and a documented appointment. Most residents have none of those.
• One correction to a claim in circulation: no well sourced case of a misidentified delivery driver could be found. Advocacy materials reference the scenario. Do not assert it.

Accuracy, measured where it counts

• Flock’s technical specification, quoted: “In optimal deployment conditions, we measure a greater than 99 percent capture rate in clear weather and rain, greater than 96 percent OCR accuracy and greater than 97 percent plate-state accuracy.” Note the qualifier. Flock’s marketing has separately claimed it accurately captures 93 of every 100 plates. Those two numbers have never been reconciled publicly, and asking the vendor to reconcile them is a fair and revealing question.
• The per-scan rate is the wrong measure. The alert-level rate is the right one, because alerts are what send an officer to a car. In Roseville, California, of 1,427 stolen vehicle and felony alerts across 2023 and 2024, 71 percent involved an incorrect plate read. Roseville also reported no arrests from Flock scans alone after spending roughly $450,000, said it had reported problems to Flock dozens of times, and disputed Flock’s claim that performance had significantly improved. Flock attributes the result to older hardware and non-standard mounting unique to that city.
• LAPD Inspector General, July 2026: a 32.3 percent error rate, and 161 false stolen vehicle alerts in two months, each sending officers to an innocent driver. A Vallejo audit found 37 percent of fixed reader hits were misreads.
• Why the error rate turns into trauma rather than an apology. LAPD, like many departments, treats a suspected stolen vehicle as a high risk stop requiring backup, air support, a supervisor and ordering the occupant out of the car. Its own watchdog warned the inaccuracies create “substantial legal and financial liability concerns” and could “potentially create a confrontation with serious consequences.”
• How little of the data is ever relevant: the California State Auditor found LAPD held 320 million plate images of which only 400,000 produced an immediate hotlist match. Ninety nine point nine percent of the stored images were of vehicles on no list at all.
• The honest caveat, which a vendor will raise: a misread only becomes a wrongful stop if the mistaken string happens to collide with a plate on a hotlist, so raw misread volume overstates the risk. Two things make that reassurance thin. Character confusion errors are not random, they cluster on visually similar glyphs, which narrows the space of wrong answers and raises collision odds above chance. And the Roseville data measures the collision layer directly.
• On the 99 percent accuracy standard raised locally: no state imposes an accuracy standard on ALPR vendors and North Carolina imposes none. If the Council wants one, it has to be a contract term with a remedy attached, such as withheld payment or termination on failure to meet a measured alert-level accuracy threshold. Nothing in the standard agreement provides it.

Insider misuse: the best documented harm in the entire record

• The Institute for Justice’s abuse database listed 197 incidents as of September 3, 2026. Tabulated directly for this briefing, 89 records are non law enforcement use, 55 are stalking, 28 are technical error and 21 are other governance failures. That means 75 percent of documented plate reader abuse is deliberate human misuse and only 15 percent is a machine getting a plate wrong. Section 5 breaks the whole database down by category, year, state, vendor and outcome, because the distribution is more persuasive than any single case.
• Individual cases, with outcomes: a Lufkin, Texas officer ran more than 10,000 lookups and was indicted on 100 felony counts. A Katy, Texas officer ran more than 2,000 and was fired and indicted. A Richmond County, Georgia deputy ran 1,639. A Sedgwick, Kansas officer ran more than 200 and resigned. A Milwaukee officer searched a girlfriend’s plate 124 times. A Washington Post investigation found roughly 46 to 50 officers accused of using Flock to track women, including a Florida officer who made 700 unauthorized inquiries tracking an estranged wife.
• Access reaches well past patrol officers. Documented cases involve a school resource officer, a 911 director, a real time crime center analyst and a district attorney’s office employee.
• The detection mechanism is a victim, not a control. The Institute’s own summary: “Most incidents came to light only after victims reported the officers’ behavior to police, typically in a broader stalking allegation.” Not through audits. Not through the software.
• And the audits mostly do not happen. The California State Auditor examined four agencies in 2020: “None had audited searches of the ALPR images by their staff and thus had no assurance that the searches were appropriate.” The LAPD Inspector General found no thorough audit since 2022, and when asked under the state’s public records act in 2025, LAPD reported having no records of search reviews at all. Denver’s internal audit found 17 percent of searches lacked a documented case number.
• A case number field is not a control, it is a data entry field. When Flock piloted voluntary case number entry in July 2025, officers bypassed it by typing “investigation,” “crime,” “hehehe,” or bare punctuation. There is no mechanism verifying that the plate searched has any connection to the case number entered. It becomes a control only when case numbers are validated against the records management system and a sample of searches is reviewed against the underlying files.
• North Carolina is not exempt, and the state’s three 2026 records are all misuse rather than error. A Brunswick County Sheriff’s Office employee resigned in September 2026 and was placed under criminal investigation. A Mooresville officer was criminally charged in August 2026 for allegedly using Flock to track her boyfriend’s ex-wife. And in June 2026 a Charlotte-Mecklenburg officer was placed on leave and arrested for using Flock to look up an undercover investigator’s vehicle at the request of a drug suspect. That last one is an officer safety failure, not a privacy failure, and it is the answer to anyone who argues that officer safety runs in favor of these systems.

Federal and out of state access

• Illinois, August 2025. The Secretary of State’s audit found 46 out of state agencies had searched Illinois plate data in violation of state law, including 262 immigration related searches on one village’s system, and that U.S. Customs and Border Protection had accessed Illinois data. Flock’s CEO conceded: “We clearly communicated poorly. We also didn’t create distinct permissions and protocols.”
• Forest Park, Illinois. A detective investigating an arson accidentally accepted more than 200 data sharing requests, one of them from ICE. The village’s plate data flowed to federal immigration enforcement for 17 months before anyone noticed, and it was the vendor that notified the village, not the other way around. The officer faced no disciplinary action. The mayor’s account: “It wasn’t done intentionally.” This is the clearest documented instance of the accidental consent failure mode, and it is the same mode Apex experienced in miniature.
• Johnson County, Texas. A sheriff’s deputy ran two searches, the larger of which reached 83,345 cameras across 6,809 networks. The reason logged in the system was “had an abortion, search for female.” Flock publicly called the reporting “false,” “misleading” and “clickbait,” and said the woman was being searched for as a missing person. The Electronic Frontier Foundation then obtained the records, which show a death investigation in which the detective consulted the district attorney about charging her. The vendor’s public account and the primary records do not match.
• Washington State. A University of Washington human rights center review of 35 responding agencies found at least eight had explicitly enabled sharing with Border Patrol, and thousands of searches ran across at least ten more agencies without any authorization, with search terms including “immigration,” “ICE” and “criminal alien.”
• Pflugerville, Texas, August 2026. A resident’s public records request revealed outside agencies had searched the city’s camera data approximately 1.6 million times between February 16 and August 15, 2026. The Council terminated the contract unanimously and disabled 28 cameras. A council member’s reaction: “We need to get them the hell out of here.”
• Note the pattern in all four: the town found out afterward, from a records request, an audit, a journalist or the vendor. Apex’s June 2026 disclosure belongs in the same list.

Security

• November 2025. An independent researcher published vulnerabilities in Flock’s plate reader, audio detection hardware, compute box and mobile applications, registered as formal CVEs. Flock confirmed the vulnerabilities exist while arguing that exploitation would require physical access and internal hardware knowledge, and said no customer action was required.
• December 2025 and January 2026. At least 60 Flock Condor cameras were left reachable on the open internet, where anyone could watch live feeds, download roughly 30 days of archived video, and change settings, including deleting video. Flock called it “a limited, isolated configuration issue” affecting a small number of devices. Most affected municipalities declined to say whether they were among them. Flock’s CEO characterized the reporting as a “coordinated attack” from “activist groups.”
• Longer precedent in the sector: in 2015 researchers found plate reader cameras in Louisiana and Florida exposed without authentication, streaming live video and plate captures. In 2019 a Customs and Border Protection subcontractor, Perceptics, moved federal traveler data onto its own network without authorization, was breached, and 184,000 traveler images ended up on the dark web; the DHS Inspector General found the agency “did not adequately safeguard sensitive data on an unencrypted device.” In 2024 Michigan State Police cyber investigators found seven vulnerabilities in Motorola’s plate readers, including a single hardcoded WiFi password identical across all cameras.
• What is not established, and should not be claimed: there is no confirmed case of a malicious actor stealing a bulk ALPR plate database. The documented pattern is exposure and access control failure, not proven mass theft. Saying it that way is both accurate and serious enough.

Scope creep: what outlives the vote

What a town approves in year one is not what it operates in year three. Every item below was added to existing ALPR deployments elsewhere, after the original approval.

• Video with AI natural language search. Still cameras became video cameras with live feeds and searches phrased in plain English, the ACLU’s cited example being a “landscaping trailer with a ladder.” This makes occupants and bystanders capturable and searchable, not just plates.
• Vehicle Fingerprint, which finds cars with no plate read at all, searching on decals and bumper stickers, roof and rear racks, and temporary tags, with a function for finding multiple vehicles believed to be traveling together. A bumper sticker is expressive political speech, and this turns it into a searchable index term. The traveling-together function is associational mapping. This capability produced the San Diego wrongful jailing described above.
• Raven audio detection, announced October 2025: microphones above city streets detecting gunfire and human sounds including screaming. An ALPR contract that becomes an audio surveillance contract is the clearest single scope creep risk to name in a vote.
• Drones. Flock partnered with Aerodome in July 2024 and acquired it for $300 million in October 2024, so a plate alert can now cue a drone launch. By late 2025 the drone product was being marketed for tracking shoplifters.
• Flock Nova, which links plate data to commercial people-lookup services, letting an investigator jump from a plate to a person. Reporting indicated internal concern that Nova incorporated data from hacks and breaches; Flock later decided not to include breach data. EFF’s reaction to the design: “an officer having someone’s entire online persona one click away after a license plate reader scans their plate is an open invitation for police retribution.”
• The Business Network, which lets private companies share vehicle hotlists with each other. The council question this raises: does the Town’s policy contemplate a private company placing a vehicle on a list that generates a police alert?
• What is not supported, and should not be said: no ALPR vendor ships facial recognition as a product feature. The supportable claim is the one above, that video and AI search make occupants and bystanders capturable and searchable.
• How it arrives, documented. San Diego has a surveillance ordinance requiring Council and privacy board approval for new surveillance technology. In December 2025 its police department added Flock Nova on a $0 first year pilot without returning to Council, on the theory that Nova is exempt because it only analyzes data from already approved technology. The price after year one is $230,000 annually.

5The abuse record, counted

Section 4 described the harms case by case. This section counts them. The Institute for Justice maintains the only comprehensive public index of documented plate reader abuse in the United States, and the tabulation below was taken directly from that database on September 3, 2026. It is the single most useful document available to a council, because it converts an argument about anecdotes into a distribution.

Violations by category

Violations by year

• 110 of the 194 records, 57 percent, are from the first nine months of 2026 alone, against 84 for the five years before that.
• Read the trend carefully, because there are two possible explanations and the honest answer is both. Deployment roughly tripled over this period, so more cameras produce more misuse. And audits, records requests and journalism increased sharply in 2026, so more of the misuse that was already happening became visible. The database cannot separate the two. What it does establish is that the visible rate is rising fast, and that a town renewing in January 2027 is renewing into a period when this conduct is being found and charged rather than overlooked.

Violations by state

Which system was involved

• Flock: 182 of 194 records, or 94 percent. The remainder are one Rekor incident (Fayetteville, North Carolina), one NDI incident (Charlotte, North Carolina), one Guardian incident, and eight where the system is unspecified.
• State that figure carefully. Flock’s share of the index tracks its share of the market and its share of press coverage. It is not a controlled comparison between vendors, and it is not evidence that a different vendor’s system would produce less misuse. What it does show is that the incident record for the alternative vendors is close to empty because there is almost no deployed history to draw on, which is the same point Section 12 makes about Axon.

What happened to the people who did it

Counted from the database’s own incident descriptions. A single record can carry more than one outcome, so these do not sum to 194.

• This is not a ledger of policy violations. It is substantially a criminal ledger. Roughly one in four records involves an arrest, a charge or an indictment, and more than a quarter of the people involved lost their jobs. A system whose misuse is prosecuted as a crime is not a system where an internal policy and an annual self-audit are proportionate safeguards.

The largest individual cases

• The number to sit with is 10,000. One officer, one system, more than ten thousand searches before anyone noticed. Nothing in the software stopped him, and nothing in an annual self-audit would have.
• And note the range of consequence for comparable conduct: 165 searches of a single plate produced a ten day suspension in one town, while 166 searches produced a firing in another. There is no consistent standard, which is another way of saying the standard is whatever the department decides after the fact.

The five North Carolina records

• Three of the five are from 2026, and all three of those are misuse rather than error.
• The Charlotte case from June 2026 is the one to put in front of anyone who argues that officer safety favors the cameras. The system was used to expose an undercover investigator’s vehicle to a drug suspect who asked for it. That is not a privacy harm. It is an officer safety harm created by the existence of an easily queried location database inside the department. Any argument that plate readers protect officers has to account for that entry.
• None of these five involve Apex, and that is the point. Two are in Charlotte, one in Mooresville, one in Fayetteville, one in Brunswick County. Each was discovered by an audit, an arrest or a lawsuit. Apex has run no published audit of its own searches.

The 21 governance failures, which are closest to what Apex experienced

The database’s “other misuse” category is the one that matches the Town’s June 2026 disclosure. Every entry below happened at an agency that had a policy, and the policy did not prevent it.

• Unauthorized access by outside agencies. Pflugerville, Texas covered and shut down its cameras in August 2026 after a resident’s records request showed hundreds of outside agencies had unauthorized access. Story County, Iowa discovered other agencies had obtained access to its Flock data without permission. San Francisco had more than 1.6 million searches of its data run by out of state authorities.
• Federal and immigration access against the agency’s own policy. Richmond, Virginia shared Flock data with the FBI in violation of a departmental ban on federal sharing, and separately granted system access to a federal Bureau of Alcohol, Tobacco and Firearms analyst who then ran immigration enforcement searches. Sparks, Nevada receives “immigration violation” hot list alerts from ICE despite an internal policy prohibiting immigration enforcement use. Eagan, Minnesota’s suburban data was searchable by out of state agencies for civil immigration enforcement despite a state law limiting it to criminal cases. Dayton, Ohio’s system was accessed by more than 140 agencies nationwide for immigration searches that city policy did not allow. An Illinois Secretary of State audit found Customs and Border Protection had obtained access to Illinois data in violation of state law.
• Retaliation and political surveillance. Multiple Waukesha County, Wisconsin deputies ran more than 100 searches on one man’s car, allegedly in retaliation for a complaint he had filed against a deputy. An Akron, Ohio analyst and officer tracked a prominent activist critical of police, logging the reason as “activist recording,” which violated department policy. Lenexa, Kansas police allegedly tracked the movements of a writer who had published a column critical of them. An Electronic Frontier Foundation investigation found agencies nationwide logged hundreds of searches related to political protests between December 2024 and October 2025.
• Data leaving the system. A Sandy Springs, Georgia reserve sergeant was fired for transferring Flock data to a private company he also worked for. A Louisville, Kentucky detective was disciplined for sharing his Flock login with a federal Drug Enforcement Administration agent. A San Francisco officer obtained Flock images of a relative’s stolen car and posted them to Instagram. A former West Chester, Ohio officer was sentenced to five years of probation for tampering with Flock data and images.
• Two entries that speak for themselves. In Millcreek, Utah, officials found a Flock camera on city property and removed it because they could not determine who had installed it or who had access to it. And a Houston Chronicle analysis found that the overwhelming majority of Flock searches by Houston Police Department employees offered minimal or no justification at all.

The 28 error records, in aggregate

• 13 of the 28 explicitly describe officers with guns drawn. Six involve someone being jailed, including 13 days in Volusia County, Florida and nearly a month in San Diego. Four involve children present, including a three year old watching her grandparents held at gunpoint in Morristown, Tennessee and a six week old left alone in a car in Sherwood, Arkansas.
• The failure modes divide almost evenly between the machine and the paperwork. Character misreads account for roughly half: an O read as a 0, a 7 read as a 2, a single digit in Sherwood, a temporary tag misclassified in Sedgwick County. The other half are records failures: a plate left on a hot list after the vehicle was recovered, an officer typing the wrong plate into a hot list, an alert misinterpreted. Four separate Colorado records describe the same failure, a driver repeatedly pulled over because an officer put the wrong plate on a hot list and it could not be removed.
• One entry in this category is not an individual stop. It is the Roseville, California finding: 1,427 alerts over two years, of which 71 percent rested on an incorrect plate read. It appears in the same list as the wrongful stops because it is the statistical version of them.

6What the Town’s own records show

Everything in Sections 3 through 5 is national. This section is not. It is drawn from the Town of Apex’s own email, released under the North Carolina Public Records Act, together with a spreadsheet the Police Department prepared for its own Chief. No advocacy group, no vendor and no news outlet stands between these documents and the reader. They are the most authoritative material in this briefing, and they are also the least flattering to every party involved, including the residents pressing the issue.

6.1 What the Department itself says the system has produced

On April 9, 2026 the Department’s Compliance Specialist sent the Chief of Police a message titled “ALPR Notes,” attaching a spreadsheet named Flock Success. It was assembled by the Department, for the Department, to document the system’s value. It is therefore the strongest version of the case for the cameras that exists in Apex, and it should be read as such.

• Eleven named incidents. Recovered stolen vehicles, a warrant service originating from a Raleigh officer’s Flock hit, a domestic violence protective order violation, a flee-to-elude stop on US 64, a stolen U-Haul recovery that turned up a second stolen vehicle, and a traffic stop producing methamphetamine and fentanyl charges. These are real outcomes and Section 3’s argument applies to them in full.
• Ninety one calls for service carried the words Flock, LPR or ALPR in their computer aided dispatch notes between September 1, 2025 and the date of the email, spread across 28 different call types.

The 91 calls, by type

• The mix is worth noticing. Roughly a third of the 91 calls are property offences, which matches Apex’s crime profile. Another one in eight are missing persons, welfare checks, suicide threats or a deceased person, which are not crimes at all. The humanitarian uses are real, and they are also the uses that a much narrower system, with far shorter retention, would serve equally well.

6.2 In April 2026 nobody could say how many cameras Apex had

Section “Where Apex stands today” notes that the number of live cameras has been reported publicly as seven, eight and ten. The records explain why that was never resolved from the outside. It was not resolved on the inside either.

• April 10, 2026. Flock’s account executive wrote to the Deputy Chief: “Apex currently has 16 active deployments, each showing varying installation dates.”
• April 15, 2026. The Deputy Chief replied: “you indicate 16 active deployments. Can you confirm what this means? We don’t currently have 16 cameras deployed, but that’s how I read your sentence.”
• April 16, 2026. The account executive: “That is the number that I was communicated to me but let me double check. It doesn’t sound right to me after our conversation on Tuesday.”
• The request that started it. On April 10 the Deputy Chief had asked the vendor how to see, on his own end, the total number of cameras deployed and their installation dates. He was, in his own words, “working on a public presentation as we reaffirm the benefit of these are to our community.”

6.3 The system was being expanded while the contract was under review

• March 10, 2026. The Deputy Chief asked staff to route an amendment converting three of the five Flex cameras to fixed units: “Previous administration had us deploying 5 flex cameras and this is ridiculous.” The Assistant Town Attorney had reviewed it with no concerns.
• The money. Installation runs $650 per camera, $1,950 for three, and the Deputy Chief noted it “will have to cover that from another budget line.” A $1,500 credit applies because fixed Falcons cost less than Flex units. Staff had to ask twice why the figures did not net out. The contract sits under Capital Equipment in the budget.
• April 17, 2026. Staff wrote to Flock describing two documents in flight at once. The first was that amendment. The second, called an addendum internally, was a Law Enforcement Agreement “Adding additional 10 cameras to existing contract.” Staff noted the quote had expired and still showed discounts, and asked for an updated version for submission.
• The clock. The same day, the Deputy Chief wrote: “we were just notified PO deadlines are in 10-days, so routing contracts for approval is super tight… They are cutting off POs a month ahead of what we expected.”

6.4 A reported camera compromise that was never resolved

The most serious document in the set is a May 1, 2026 email from the Town’s Information Technology department to the Chief of Police.

• The origin. Residents contacted the Council saying they believed Flock cameras could be hacked. A senior Town official, identified in the record only by first name, asked IT to look into it. IT delivered a written briefing in February 2026.
• The meeting. A meeting including that official, IT and the Chief ended with an agreement that the Chief would supply Flock’s contact details so the vendor could reach IT directly, and that IT would continue its investigation once that contact occurred.
• April 14, 2026. The Chief wrote to IT: “you stated that someone you knew had successfully hacked into one of our deployed ALPR cameras. You cited lack of a key security patch as the likely reason this was able to be done. In an effort to reconcile what our vendor is telling me, and what happened with this hack, I need specific information. Namely, I need to know which of our cameras was accessed, when, and by what means.”
• May 1, 2026. IT’s answer: “To date, we (IT) have not been contacted by Flock.” IT confirmed it had done no further work, having been told to wait for the vendor, and added: “It’s also possible that, given the three months that have passed since the original concern was raised, Flock may have already addressed or resolved any issues on their end. However, we have not verified this.”

6.5 Federal and state access requests, in April

A short internal thread between the Department’s Criminal Intelligence Analyst and the Deputy Chief tracks outside agencies asking to reach Apex’s Flock data.

• April 8, 2026. “I wanted to let you know there is a state agency requesting Flock access.”
• April 13, 2026. “There is another state agency request pending in Flock.”
• April 15, 2026. “Two federal agencies are requesting access in Flock.”
• What the record does not show. The Deputy Chief’s replies are “Done. Thank you.” and “Thank you.” Nothing released indicates which agencies, what standard was applied, who held the authority, whether any request was granted, or whether a record of the decision exists.

6.6 What residents compiled, and what the Council already has

Two resident-prepared documents are in the file, one from July 2026 and one from August 2026. Both are advocacy, and both are footnoted to primary sources including IPVM, a United States Senator’s letter to the Federal Trade Commission, University of Washington researchers and Flock’s own executive emails. The claims below are reproduced as resident-compiled and individually sourced, not independently verified for this briefing. Every one of them is checkable, and the Council should require staff to check them rather than dismiss or adopt them.

• Contract terms changed under the Town. In February 2026 Flock’s terms are reported to have removed the line “Flock does not own and shall not sell Customer Data” and added the word “perpetual” to the licence the customer grants Flock to use customer data. Terms are now linked online rather than included in the signed contract, and the current order form voids previous agreements.
• Audit capability was reduced. In December 2025 Flock is reported to have told customers that Network Audits “will no longer include officer name, specific plates searched, vehicle fingerprint, and open text search reason.” If accurate, Apex lost the ability to verify that agencies it shares with are honouring the agreements they signed.
• The sharing network grew fast. APD was sharing with 994 organizations on March 31, 2026, described as a 20 percent increase in four months, and with roughly 1,192 by mid-2026. Twenty seven officers and civilian employees at fifteen agencies holding access to Apex data are reported to have been investigated or arrested for Flock misuse.
• Two North Carolina agencies with access are now bound to ICE cooperation. Senate Bill 153 requires four statewide agencies to enter 287(g) agreements. The State Bureau of Investigation and the Department of Public Safety were both still listed as having access to Apex data on the Flock transparency portal as of August 11, 2026.
• The accuracy arithmetic, if it holds, is the largest number in this briefing. IPVM’s 2022 evaluation found up to a 10 percent misread rate on Flock’s Falcon. Applied to Apex’s own 30 day unique read volume, the July memo calculates roughly 35,372 misread plates in a month, which implies the system reads on the order of 354,000 unique plates every 30 days in a town of about 76,000 people. Confirm that read volume against the Department’s own portal before using either figure. If it is even approximately right, it is the most important operational fact about the program, and it has never been presented to the Council.
• A resident wrote to the Chief on June 25, 2026 asking about external sharing settings and reported no reply by the time the August document was written.

6.7 Records requests, and what the redactions cover

• Public records request PUPR-2026-092 was logged on February 17, 2026. On February 26 staff recorded the requester following up and could not give a target completion date. On March 19 the Public Records Coordinator was still asking the Chief and Deputy Chief whether records existed answering points 5 through 10 of the request. The thread continues to April 2.
• Nothing in that thread suggests obstruction. It suggests a small clerk’s office and a legal review queue absorbing a detailed request. It is also six weeks in which a resident asking a specific question got no specific answer.
• The redactions land on the substance. In several of the released threads the routing, the signature blocks and the courtesies survive while the operative reply is blacked out. That is often lawful. It also means the public record, as released, frequently shows a question and not its answer, and this briefing has not filled those gaps with inference.

6.8 Two smaller findings worth a line each

• A private country club feeds the police network. On April 3, 2026 Flock notified Apex PD that MacGregor Downs Country Club “has just shared their cameras with Apex NC PD, granting you and your organization the ability to search footage from their Flock Safety cameras.” This is the private-to-public data flow described in Sections 1 and 2, operating in Apex today, and it arrived by marketing email rather than by any Town decision.
• Competitors are working the decision. A rival vendor cold-emailed the Deputy Chief repeatedly through the spring, opening with “If a license plate reader rollout gets denied at council, what becomes Plan B?” The Council should assume every option in front of it, including the alternatives, is being actively sold.

7The legal theory, and why nobody voted

Two separate questions get confused constantly. The first is whether plate readers are constitutional. The second is whether anyone has to ask permission before installing them. Today the answer to the first is unsettled and moving, and the answer to the second, in North Carolina, is no.

What the vendors argue, and it is not weak

• Katz v. United States (1967). “What a person knowingly exposes to the public is not a subject of Fourth Amendment protection.” A plate is issued by the state, mounted on the outside of the car and required by law to be visible, so no expectation of privacy in it is objectively reasonable.
• United States v. Knotts (1983). “A person traveling in an automobile on public thoroughfares has no reasonable expectation of privacy in his movements from one place to another.” If an officer may follow a car without a warrant, a camera doing the same thing at one point adds nothing. But read the reservation: the Court said that “if such dragnet-type law enforcement practices as respondent envisions should eventually occur, there will be time enough then to determine whether different constitutional principles may be applicable.” Knotts answered the single camera question and expressly declined the dragnet question.
• Cardwell v. Lewis (1974) and New York v. Class (1986). “The exterior of a car, of course, is thrust into the public eye, and thus to examine it does not constitute a search.” These cases are about the visibility of an identifier. Neither addresses databases or aggregation.
• Plain view, as Flock states it: “License plates are visible on public roadways; cameras simply capture what’s already exposed.”
• The third party doctrine (United States v. Miller, Smith v. Maryland). Where a private vendor holds the images, police obtaining them is characterized as a records request rather than a search. This is now the weakest limb of the argument, for reasons in the next subsection.
• The vendor’s own brief is the most useful document here. Flock’s published Fourth Amendment position distinguishes the adverse cases on the grounds that its cameras produce “sparse, momentary observations” rather than “comprehensive, long-term, always-on logging,” and it treats retention limits as the constitutional linchpin, arguing short retention prevents a mosaic from forming. Camera density and retention length are precisely the two variables a Town Council controls. A council that shortens retention and caps camera count is strengthening the legal position the vendor itself asserts. A council that expands the grid is eroding it.

What cuts the other way

• United States v. Jones (2012). The judgment rested on physical trespass, which does not transfer to pole cameras. The concurrences do, and they represented five justices. Justice Sotomayor: GPS monitoring “generates a precise, comprehensive record of a person’s public movements that reflects a wealth of detail.”
• Carpenter v. United States (2018). Individuals have a reasonable expectation of privacy in “the whole of his physical movements,” and acquiring historical cell site location is a search. Read the limits, which advocates on both sides omit: the decision is “a narrow one,” and it does not “call into question conventional surveillance techniques and tools, such as security cameras.” That clause is cited in every single ruling upholding ALPR.
• Leaders of a Beautiful Struggle v. Baltimore Police Department (Fourth Circuit, en banc, 2021). This is binding precedent in North Carolina. Baltimore’s aerial surveillance program was unconstitutional without a warrant “because the AIR program enables police to deduce from the whole of individuals’ movements.” It is the most challenger friendly circuit precedent in the country, and it governs here.
• Chatrie v. United States, decided June 29, 2026. A government request to Google for location history is a Fourth Amendment search. Three features matter. It rejected a duration threshold, holding that a request for any amount of location history is a search and refusing a grace period for short retrievals, which undercuts the argument that a 7, 21 or 30 day window is short enough to escape scrutiny. It narrowed the third party doctrine, distinguishing the older cases as involving “limited types of personal information.” And it did not decide the particularity question, which was remanded to the Fourth Circuit.
• Do not overstate Chatrie. It concerns a government demand to a private company for records generated by a device the target chooses to carry. A town owned pole camera photographing a public street is meaningfully different, and Carpenter’s security camera carve out is still on the books. The genuinely open question is the case in the middle: a vendor hosted, nationally searchable plate database, which resembles Chatrie’s posture far more than it resembles municipal CCTV. No court has resolved that.

Where the ALPR cases have actually landed

• Every ALPR specific decision located for this briefing came out for the government on its facts. No federal appellate court and no state supreme court has held that a municipal ALPR network requires a warrant. Anyone telling this Council that the courts have struck this technology down is wrong.
• And the leading decision is expressly contingent. Judge Mark Davis wrote in Norfolk that plate reader surveillance “could become too intrusive and run afoul of constitutional privacy standards at some point,” but that “at least in Norfolk, Virginia, the answer is: not today,” and that “as the number and capabilities of ALPR cameras expand, the constitutional balancing could conceivably tip the other way.”
• The single most important legal fact for Apex. Schmidt is on appeal in the Fourth Circuit, docket number 26-1227, the circuit that governs North Carolina. That is the same court that decided Leaders of a Beautiful Struggle and the same court now holding Chatrie on remand. The district court in Norfolk ruled five months before Chatrie was decided, without the benefit of a Supreme Court decision that rejected duration thresholds and narrowed the third party doctrine. The Fourth Circuit will be the first federal appellate court to apply Chatrie to a plate reader network.
• The practical consequence: the governing law in North Carolina on municipal ALPR is likely to change within twelve to twenty four months, in a direction nobody can currently predict, and the Town will not control the timing. Any contract signed now should be terminable without penalty, and any retention setting adjustable without penalty. That is not an ideological demand. It is ordinary risk management on an unsettled question.
• Also pending: the Kansas Justice Institute sued the City of Wichita in July 2026, arguing that plate readers make a license plate “the functional equivalent of a GPS tracking device.”
• On the due process principle raised in local discussion, stated accurately. Coolidge v. New Hampshire (1971) suppressed evidence taken from a car in the murder of a fourteen year old girl, because the warrant process used was invalid. It is not a holding about cameras. It is the clearest illustration of the principle that the emotional weight of a crime does not validate an unlawful method of solving it, which is the correct answer to any argument that begins with a murder and ends with a request to skip a procedure.

North Carolina law: Article 3D of Chapter 20

• A correction worth having ready. At least one widely circulated fifty state comparison lists North Carolina’s retention as “per policy” with no statutory cap. That is wrong. The 90 day cap is statutory. If a vendor or consultant tells the Council North Carolina has no retention limit, they are mistaken.
• The gap is the whole point of this section. North Carolina regulates how long data is kept far more tightly than it regulates who may look at it and why. Article 3D governs retention and preservation, not querying. Inside the 90 day window an officer may query a resident’s location history with no warrant, no case number requirement, no logged justification and no independent audit. The statutory audit is a self audit, written and conducted by the agency, reported to the agency.
• What tighter states do, for comparison. Virginia (2025) requires 21 day retention, a two year audit trail, per query logging of date, plate, specific purpose including offense type, case number and username, court order or subpoena for sharing, annual public reporting, and independent verification of an alert before a traffic stop. Maine requires 21 days. New Hampshire keeps non-hits for roughly three minutes. The Policing Project’s 2026 fifty state scorecard found 24 states have ALPR legislation, 26 have none, and not one state satisfied all seven of its criteria.
• And North Carolina moved in the opposite direction this summer. The 2026 Appropriations Act, Session Law 2026-41, signed July 7, 2026, made the State Bureau of Investigation’s highway plate reader program permanent on NCDOT owned or maintained rights of way, roughly 80 percent of state roads, with the SBI acting on behalf of federal, state and local agencies. The program had roughly 140 cameras and 32 participating agencies as of March 2026, with over 150 million scans recorded, and the SBI’s stated goal is cameras at every entrance to and exit from the state. Reporting requirements were narrowed in the permanent version, which no longer requires public disclosure of total plate captures or data preservation instances. A town that declines its own cameras may still be covered by state cameras it has no vote over. A town that keeps its cameras may be feeding that network.

Consent of the governed: there is no requirement

• No North Carolina statute and no federal law requires public notice, a public hearing or a referendum before a municipality installs plate readers. Article 3D requires a written policy. Not a hearing, not a vote, not notice. Residents of Moore County were told exactly this at a special meeting in August 2026: the sheriff could adopt cameras without a vote, referendum or ballot measure.
• What does apply is general procedural law: the Open Meetings Law requires that a meeting where the Council acts be noticed and open, and the budget hearing under Chapter 159 is the one guaranteed public comment opportunity if ALPR appears as a line item, which in practice means one hearing on a document hundreds of pages long.
• One structural advantage Apex has: in several North Carolina counties the sheriff, an independently elected constitutional officer, made this decision alone, and commissioners said they do not set law enforcement operational policy. A town council controls its police department’s budget and can condition appropriations. Municipal leverage here is real in a way county leverage often is not.
• California is the counterexample. Civil Code section 1798.90.55(a): “A public agency that operates or intends to operate an ALPR system shall provide an opportunity for public comment at a regularly scheduled public meeting of the governing body of the public agency before implementing the program.”
• Community Control Over Police Surveillance ordinances exist in somewhere between 15 and 26 jurisdictions depending on which count you use, including Seattle, Oakland, San Francisco, Berkeley, Davis, Boston, Cambridge, Somerville, Nashville, Madison, Grand Rapids and New York City. They require a published surveillance impact report and use policy, a public notice and comment period (Davis requires 30 days before the hearing), an affirmative council vote, annual reports, and annual council review with authority to modify or discontinue. The strongest versions add a private right of action, damages, attorney’s fees and evidentiary exclusion. No North Carolina jurisdiction has one.
• Referenda are required nowhere. Conroe, Texas saw a citizen effort to put Flock cameras on the November ballot, and the City Council voted to keep it off. An Arizona bill introduced in March 2026 would require supermajority voter approval before a city deploys surveillance technology; it has not been enacted.

How a vote gets avoided, and what to do about it

• The threshold analysis is not the important one in North Carolina. Formal bidding starts at $90,000 for equipment and informal bidding at $30,000, and G.S. 143-133 provides that “No bill or contract shall be divided for the purpose of evading the provisions of this Article.” But ALPR is procured as a service subscription, and service contracts are not subject to North Carolina’s competitive bidding statutes at all, at any dollar figure. Whether this Council votes is therefore set by the Town’s own purchasing policy and the manager’s delegated signature authority, not by state law. That is the single most actionable finding in this briefing: if the Council wants a guaranteed vote on surveillance technology, it has to write that requirement itself.
• Per camera annual pricing makes deployments easy to size to an approval level. In Pittsboro, North Carolina, the price was $3,750 per camera per year plus $650 installation per unit. At that rate 24 cameras is $90,000 and ten cameras is $37,500, which under a typical delegation a manager signs. Multi year totals that would clear a threshold are commonly presented as annual figures.
• The one clearly documented threshold case, stated with care. In Anoka County, Minnesota, the sheriff signed a $174,999 Flock camera and drone contract on June 30, 2026, against a board approval threshold of $175,000. No board vote occurred, and a commissioner said she was neither consulted nor informed until after the contract was final. The dollar figures are verified. Intent is not, and the reporting itself says the evidence does not establish that the figure was set to bypass oversight. Present it exactly that way. It is powerful enough without embellishment, and a council member who overstates it will be corrected.
• The most effective vote avoidance structure is the free pilot, and Denver is the case study. Denver’s City Council unanimously rejected a $666,000 two year Flock extension in May 2025, citing Fourth Amendment concerns, federal access fears, constituent opposition, and an internal audit finding officers failed to document case numbers in 17 percent of searches. In October 2025 the mayor extended the contract at no cost through March 2026, and because there was no cost, council approval was not required. A council majority publicly called the arrangement undemocratic. A $0 pilot defeats both the dollar threshold and the cost objection, while establishing operational dependency before any priced renewal reaches a council.
• A claim to drop. An advocacy contract tracker states that after Denver’s council rejected the renewal, the city signed a $498,500 extension, $1,500 under the threshold requiring a council vote. That figure could not be corroborated and conflicts with contemporaneous local reporting describing a no cost extension. Do not use it. The verified version makes the same point without exposure.
• The drafting response, five clauses. Require Council approval for any new surveillance capability, defined to include new analytic products and new data sources applied to existing collection, not merely new hardware. Make $0 acquisitions, pilots, trials, grant funded deployments and donations subject to the same approval as purchases. Require approval for amendments, renewals and extensions, not just initial execution. Require agenda materials to disclose total contract value over the full term, per camera pricing and the cost of every optional module. And set the trigger in local policy, because state law will not supply it.

8What this could cost Apex

The vendor’s maximum contractual liability on this contract is roughly $35,875. The benchmark settlement for a family held at gunpoint over a misread plate is $1.9 million. Everything in between is the Town’s.

What has actually been paid

• The going rate for comparable claims where ALPR was not the trigger, useful because it shows the category price rather than the technology price: $225,000 in Denver in 2026 for two people detained at gunpoint after a human plate misidentification, $250,000 in North Myrtle Beach, $350,000 in Columbus.
• A defensible planning range: roughly $225,000 to $500,000 for a single plaintiff, and $1.9 million as the benchmark where children are involved. Two or three claims in the documented range, plus plaintiffs’ attorney fees and the Town’s own defense costs, reaches seven figures without anything unusual happening.
• One number to avoid. A hypothetical circulating locally puts a single resident’s claim at $4.6 million. Nothing on the public record supports a figure that size for this fact pattern. The documented range is more persuasive precisely because it is documented.

Why the Town holds the risk, not the officer and not the vendor

• Officers are protected and the Town is not. Individual officers get qualified immunity unless the right was clearly established, and after the Norfolk ruling it is not, so Apex officers are currently well protected on the Fourth Amendment theory. A municipality gets no qualified immunity at all (Owen v. City of Independence, 1980).
• Under Monell, the Town is liable for its own policy, and the Council is the final policymaker. That is the structural trap. An individual officer’s misread is a bad-hit case with immunity available. A Council vote to deploy the system, and the policy adopted under it, is Town policy, which is exactly what Monell reaches. Deployment scope, retention length, whether interstate and federal sharing is enabled, and whether independent visual verification is mandatory before a stop are all Council level choices, not officer discretion.
• Fee shifting is what makes small claims worth bringing. Under 42 U.S.C. 1988 a prevailing plaintiff in a civil rights case recovers attorney’s fees; a prevailing municipality essentially never does. A plaintiff who wins one dollar in nominal damages can recover six figures in fees. For a town this size, fee exposure on a modest claim frequently exceeds damages exposure.
• North Carolina immunity does not save the Town on the claim that matters. Under G.S. 160A-485, buying liability insurance is itself the waiver of governmental immunity, and no judgment may exceed policy limits on a claim that would otherwise have been immune. But state immunity statutes cannot defeat a federal civil rights claim. The Section 1983 exposure, which is the Aurora and San Francisco fact pattern, plus fees, sits entirely outside that cap. The Council should not treat the insurance policy limit as its ceiling.
• Being warned changes the legal posture, and this is worth understanding rather than fearing. City of Canton v. Harris permits municipal liability where a policy choice reflects deliberate indifference to a known risk of constitutional violation. Once the minutes show this Council was told about a 32 percent alert error rate, 31 documented gunpoint stops and a $1.9 million settlement, “we did not know” is unavailable, and plaintiffs’ counsel will attach the minutes to the complaint. That is not an argument for not learning. It is an argument for recording, in the same minutes, the mitigation adopted in response: mandatory independent visual verification before any stop, no traffic enforcement use, retention at the statutory floor, sharing disabled, queries logged and independently audited. A record of warning plus responsive safeguards is defensible. A record of warning plus adoption without safeguards is the deliberate indifference fact pattern. The Town’s own insurance pool applies the same concept, instructing that foreseeability of risks must inform policy development.

What the vendor’s contract actually says

• Limitation of liability. Flock’s standard agreement caps its total liability at the fees paid in the twelve months before the act or omission. On Apex’s roughly $35,875 per year, the vendor’s entire contractual exposure against a $1.9 million claim is about 1.9 percent of the loss.
• Indemnification. In a standard form agreement reviewed for this briefing, Flock indemnifies the customer for intellectual property claims and for “any damage or injury to property or person directly caused by Flock’s installation of Flock Hardware.” A wrongful stop caused by a misread plate is not covered.
• The warranty disclaimer, quoted: Flock “DOES NOT WARRANT THAT THE FLOCK SERVICES WILL BE UNINTERRUPTED OR ERROR FREE NOR DOES IT MAKE ANY WARRANTY AS TO THE RESULTS THAT MAY BE OBTAINED FROM USE OF THE FLOCK SERVICES.” Contractually, the misread rate is the Town’s problem, not the vendor’s.
• These terms are negotiable, and one town proved it. Dunwoody, Georgia negotiated a real data security indemnity covering third party claims arising from a security incident, a clause providing that “Flock shall not use Customer Data, whether in identifiable or de-identified form, to train, fine-tune, or improve any machine learning, artificial intelligence, or algorithmic models except as expressly authorized in writing,” and insurance requirements of $5 million in professional liability and $5 million in cyber liability. Accepting the standard form is a choice, not a constraint.
• The marketing and the contract diverge on the point that matters most. Flock’s public materials say “You own your Flock camera data, and you decide who can access it.” The ACLU reports that the standard user agreement also grants Flock a “worldwide, perpetual, royalty-free” right and license to “disclose the Agency Data for investigative purposes”, operating independently of the department’s own sharing settings unless a customer negotiates it out. Boston negotiated it out. Ask for the redlined agreement, not the FAQ.

On the claim that Flock never deletes anything

• The strong version of this claim is not established and should not be asserted. The accurate version is more useful anyway: deletion is a vendor side event that the Town cannot observe, and the Town has no mechanism to confirm it happened.
• Retention is a configurable setting, and it has been set far higher elsewhere. An LAPD audit surfaced a Flock contract permitting retention for five years, sixty times the standard policy.
• Deactivation may not be self executing. Littleton, Massachusetts defunded and de-energized six cameras by town vote in May 2026. In early August 2026 police discovered Flock had restored power to five of the six without authorization. The company’s explanation was that under its standard operating procedure it restored power because it had not been told the shutdown was permanent. The police chief’s comment is the useful one: “Public safety will always be a priority for this department, but it must be pursued in partnership with the community we serve, not around it.”
• Whatever Apex decides, put it in writing with three elements: permanence, physical removal, and certified deletion of existing data, verified at the vendor’s storage environment by someone with no financial relationship to the vendor.

Insurance and staff burden, the unglamorous exposure

• Apex’s liability coverage runs through the North Carolina League of Municipalities pool, where the base package is general and auto liability plus property, and, in the pool’s own words, “all other coverages are optional.” Law enforcement liability and cyber are optional lines. Three questions for the Town Attorney and the League, before any vote: does Apex carry those optional lines, at what limits and with what retention; does the general liability form contain an invasion of privacy or statutory violation exclusion that would defeat coverage for a privacy claim; and will the pool confirm in writing that ALPR deployment does not change coverage terms, and price the change if it does. The Town’s actual limits are the single most important number this Council does not currently have.
• The pool’s law enforcement risk review program audits member departments across dozens of dimensions and ties documented compliance to maintaining insurability, checking things like whether camera system media access is controlled and limited to a supervisor and whether random supervisory review and audit occurs. Gaps become both a coverage argument and a Monell exhibit.
• Article 3D requires an annual audit and report of system use and effectiveness. Someone has to query the logs, analyze hits against errors, and write it. On the documented error rates, that report will say uncomfortable things in a public document. It is also the only way the Town gets numbers about itself, which is an argument for doing it well rather than not doing it.
• Litigation holds collide with the 90 day purge. Once a claim is threatened, the duty to preserve and the statutory retention ceiling pull in opposite directions. The Town needs a documented, tested mechanism to suspend automated deletion for specific records without breaching the retention statute.
• Public records requests on ALPR are legal review, not clerical work. Each one has to be assessed against Article 3D’s confidentiality provisions and the criminal investigation exemption. Durham has claimed entire Flock audit trails are criminal investigation records, a position that would make all usage logs confidential. A contract term requiring audit outputs to be produced in a releasable form forecloses that argument in advance.

The 2024 utility billing attack, and precisely what it proves

• Timeline. Attack window June 23 to July 2, 2024, discovered July 2, 2024, characterized by the Town as an attempted ransomware act. No group publicly claimed responsibility.
• Scope. Personal data for 22,601 people was exposed, including names, Social Security numbers, government IDs, passport numbers, financial account and card numbers, usernames and passwords, dates of birth, health insurance information and medical records, and phone numbers. That is a far wider data set than a utility billing system should hold, which suggests other systems shared the compromised environment.
• Recovery, and why it worked. The data was exfiltrated to a United States cloud storage provider, which refused to release it without a court order. In October 2024 Wake County Superior Court issued a temporary restraining order compelling full and direct access, and the Town recovered its data. Town Attorney Dan Greene: “Because the provider was subject to U.S. law and the jurisdiction of our courts, we were able to obtain a restraining order that compelled the return of Apex’s data.” That outcome depended on the holder being inside United States jurisdiction and responsive to a court order. It is a good outcome and a narrow one.
• Notification took 21 months. Discovered July 2, 2024. Victims notified at the end of March 2026. The delay has not been publicly explained.
• Operational and financial damage. Water bills as high as $1,800 against a normal bill of roughly $250, with some metered readings up fifteen to twenty times. Residents were told to cancel automatic payments. A third party firm was hired to review incident data and run simulated bills for all 30,000 customers. A new billing system went live in July 2025. Late fees and disconnections were suspended for roughly 19 months, resuming in February 2026.

• Mayor Gilbert said the gap “impacts our budget” for capital and operations. Council member Terry Mahaffey said the outstanding balances put further upward pressure on utility rates.
• No ransom was paid, and none could have been. G.S. 143-800 prohibits any state agency or local government entity from paying or even communicating with an entity that has encrypted its data and demanded a ransom. Recovery was the only available path, and it took 19 months.
• The general picture is not flattering to any small town. Of more than 3,000 local governments surveyed by the Center for Internet Security, 80 percent have fewer than five dedicated security employees and 30 percent have no formal cybersecurity process at all. Average recovery cost excluding ransom is around $1.53 million. Apex’s experience sits above that average, which means it is not the story of an unusually unlucky town. It is the story of an average town’s capacity meeting an average attack.
• Now the careful version of the argument, because the sloppy version will be attacked. The 2024 attack does not prove Flock will be breached, and Flock’s security record is a separate question addressed in Section 4. What it establishes is narrower and harder to rebut: this Town’s demonstrated capacity to hold, reconcile, account for and recover a vendor dependent data system is limited, and when something goes wrong the recovery is measured in years and millions of dollars.
• And there is a difference in kind that makes the plate database worse, not better. A billing system failure is a downtime and reconciliation problem, and Apex eventually rebuilt it. A plate database failure is a disclosure problem. There is no backup to restore, because the harm is that the movement records of 76,000 residents are somewhere else. The Town recovered its billing data because one cooperative American company held it under court order. Nobody can issue a restraining order against 7,000 agencies that already searched and exported.
• The reasonable ask: obtain the total cost of the 2024 incident internally, including forensics, the remediation firm, the replacement billing system, legal fees, notification, credit monitoring, staff overtime and written off revenue, before voting on any new data collection system. G.S. 132-6.1(c) shields incident information from public disclosure, which is why the public record is thinner than the incident’s actual scope, but nothing prevents the Council from asking.

9What the other cameras teach

Two comparisons are unavoidable. The United States already ran a twenty year experiment in automated camera enforcement, and it lost public consent even in its narrow form. Europe permits the same plate reading technology but under a legal framework no American town currently approaches.

Red light cameras: the narrow version, and it still failed

• The arc. The first American program opened in New York City in 1992. Programs peaked at 533 communities in 2012 and stand at 357 across 34 states and the District of Columbia as of September 2026, a decline of roughly a third. Speed camera programs number 374.
• Nine states now prohibit red light cameras by statute: Idaho, Maine, Mississippi, Montana, New Hampshire, South Carolina, South Dakota, Texas and West Virginia. Ten prohibit speed cameras. A larger second tier of states has no authorizing statute at all, which functionally blocks programs, so popular lists reporting thirteen or sixteen bans are conflating the two categories.
• Voters rejected them wherever they got the chance: 53 percent against in Houston in 2010, 73 percent against in Anaheim, 77 percent for a charter ban in Cleveland, 86 percent against in Sulphur, Louisiana, 76 percent in Steubenville, 72 percent in Chillicothe, 70 percent in Mukilteo and Dayton, and 54 to 65 percent in a string of other Ohio and Washington towns. One town, East Cleveland, voted to keep them.
• Why they lost, reason one: timing manipulation. Chicago shortened yellow intervals to fractions of a second below its own three second minimum, generating nearly $8 million from about 77,000 additional tickets, then stopped ticketing on the short yellows while keeping the revenue already collected. Florida’s transportation department quietly reduced its yellow light policy in 2011. When Ohio and Georgia legislated one additional second of yellow time, ticketing fell 80 percent.
• Why they lost, reason two: money. Chicago collected $284.9 million from 2.2 million tickets under one mayor. Houston collected about $44 million in four years. The counterexample matters too: Wilmington, North Carolina lost over $200,000 operating its program in 2022, and Raleigh’s netted the city nothing, which is why both ended.
• Why they lost, reason three: criminal corruption. Chicago’s contracts with Redflex reached $124 million. A retired transportation official was convicted and sentenced to ten years after taking more than $600,000 in cash amid perks valued above $2 million; the vendor’s local consultant, who had been paid roughly $2 million and passed about $570,000 in cash in envelopes at restaurants, got six months; Redflex’s chief executive was sentenced to 30 months with $2 million in restitution, plus a separate 14 month sentence in an Ohio scheme involving Columbus and Cincinnati, where the lobbyist conduit got 15 months. In Jefferson Parish, Louisiana the vendor paid a lobbyist 3.2 percent of ticket revenue and the program was suspended amid a federal grand jury inquiry.
• Two states responded by banning revenue linked vendor pay outright. California prohibits any contract providing “for the payment or compensation to the manufacturer or supplier based on the number of citations generated, or as a percentage of the revenue generated,” and Washington’s statute is equivalent. Ask what Flock’s or Axon’s compensation scales with. If it scales with cameras, reads, detections or alerts, the town has rebuilt the incentive that ended red light cameras in nine states.

North Carolina’s red light model, which is strikingly stricter

• G.S. 160A-300.1(d) lists by name the municipalities allowed to operate a program, and a town not on that list cannot lawfully run one until the General Assembly adds it by act. Apex is not on the list. Pause on that: North Carolina requires an act of the legislature before a town may photograph a car that runs a red light, and requires nothing at all before a town may photograph every car that passes.
• The state constitution’s Fines and Forfeitures Clause, with G.S. 115C-437, sends the clear proceeds of penalties to the public schools, diminished only by actual collection costs not exceeding ten percent. That requirement destroyed High Point’s program in Shavitz v. City of High Point (2006), where the vendor was taking roughly 70 percent of collections, and nearly destroyed Greenville’s in Fearrington, where the Court of Appeals found the schools effectively kept only about 72 percent of nearly $2.5 million. The state Supreme Court reversed in May 2024 on the strength of a special local act.
• The result: Charlotte ended its program in 2006, High Point litigated its away, Cary voted its out in 2012, Greenville shut down in 2022, Fayetteville let its contract lapse, Raleigh’s SafeLight expired in April 2024, and Wilmington’s expired in June 2025. As of October 2025, Greensboro was the only North Carolina municipality still using red light cameras.

What made the narrow tool defensible, in statute

• Event trigger and no faces. Washington law: cameras “may only record images of the vehicle and vehicle license plate and only while an infraction is occurring,” and “the image must not reveal the face of the driver or of passengers in the vehicle.”
• Purpose limitation. Same statute: no image or personally identifying data “may be used for any purpose other than enforcement of violations under this section nor retained longer than necessary to enforce this section.”
• A hard deletion clock tied to a case. California destroys records within six months or on final disposition of the citation, whichever is later.
• Notice before enforcement. Washington requires signs posted at least 30 days before activation; California requires 30 days of warning notices only, plus a public announcement 30 days before enforcement begins.
• Penalty design. North Carolina: a $50 civil penalty, noncriminal, no license points and no insurance points, and no owner liability if notice comes more than 90 days after the violation.

The comparison, one line at a time

The lesson is not that cameras are bad. It is that the narrow, event triggered, purpose limited, disposition deleted version of automated enforcement lost public consent anyway, in nine state bans, a dozen lost referenda and three federal corruption prosecutions. The three causes were a revenue interest, per unit vendor compensation, and unaudited discretion. A continuous plate reader program can reproduce all three, and it lacks the event trigger and the deletion clock that made the narrow tool defensible in the first place.

Europe: the same technology, a different legal floor

• The United Kingdom operates the largest system in Europe and does not hide it: over 100 million reads a day across 12,076 camera sets plus 1,878 mobile cameras, described by its own regulator as the largest non-military database in the country.
• UK retention, from the national standards: reads in national systems are deleted at 12 months; reads in local systems at 90 days; associated images are deleted the following day; and watch lists of vehicles of interest are automatically deleted after at most 56 days. Anything held longer requires case specific justification under criminal procedure legislation. There is no keep-the-hits-forever tier.
• Access is tiered by data age and offense seriousness. Up to 90 days on staff authorization; beyond 90 days requires inspector rank; a full year only for serious or major crime. Logging and audit are mandatory under national standards covering user access, searches, alterations, administrator actions, watch list loading and disclosure.
• Transparency is structural, not discretionary. The national system’s Data Protection Impact Assessment is published and kept current, most recently updated in April 2026, and there is a published subject rights notice covering access, rectification and deletion with a named point of contact.
• Enforcement has teeth. In July 2013 the Information Commissioner ordered Hertfordshire Constabulary to stop operating a seven camera cordon around the town of Royston, finding it made it “effectively impossible for anyone to drive their car in and out of Royston without a record being kept of the journey,” and that the force had failed to conduct impact assessments before installing. The Commissioner’s formulation is the plainest sentence in this whole briefing: “The use of ANPR cameras and other forms of surveillance must be proportionate to the problem it is trying to address.” The force did not appeal.
• The UK’s own regulator is not satisfied either, which is worth quoting because it comes from inside the system: at 97 percent accuracy and current volumes, “up to 2.4 million inaccurate reads per day”; the network sits in critical national policing infrastructure without an express statutory basis; and its clearest warning: “the biggest risk to ANPR as I see it is societal, it’s that people withdraw their support for it.”
• In the European Union, a plate is personal data, because it identifies the registered keeper through the vehicle register. Police use falls under the Law Enforcement Directive, which requires a legal basis in national statute, mandatory erasure time limits, logging of collection, alteration, consultation, disclosure and erasure, impact assessments where risk is high, and enforceable individual rights of access, rectification and erasure. General data protection law separately makes an impact assessment mandatory for large scale monitoring of publicly accessible areas using optical electronic devices, which describes ALPR precisely.
• Germany is the strictest, and the most instructive. The Federal Constitutional Court voided state plate scanning provisions in 2008 and again in December 2018, holding that recording and cross checking plates interferes with the fundamental rights of everyone affected regardless of whether a database match occurs. German law now requires an express statute that names the fundamental rights it restricts, protection of legal interests “of considerable weight,” a specific place and occasion such as a border area or a defined checkpoint, a documented authorization for each measure, strict purpose limitation on cross checking, and immediate, untraceable deletion of non-matches. Suspicionless, area wide, permanent scanning is unconstitutional. You may scan. You may not keep.
• Denmark is the cautionary tale for anyone drafting a retention rule. Its data protection authority ruled blanket retention of non-hits unlawful. The ministry then permitted 30 day retention inside “targeted police operations.” The national police responded by declaring all 24 fixed camera sites to be ongoing targeted operations, retaining roughly 830,000 non-hits a day, a ratio of ninety retained non-hits for every hit. A retention limit without an independent auditor is a labeling exercise.
• No European country reviewed bans the technology outright, and none retains reads for years. The UK’s twelve months is the permissive edge of the spectrum, and it arrives bundled with rank tiered access, mandatory audit, a published impact assessment and enforceable individual rights.
• The European high courts have struck down general and indiscriminate retention three times, in Digital Rights Ireland (2014), Tele2 (2016) and La Quadrature du Net (2020), requiring that retention be targeted on objective, non-discriminatory criteria and that access be subject to prior review by a court or an independent administrative body. Those cases concern telecommunications metadata, not plates, and do not bind ALPR. The reasoning transfers cleanly because a plate database is location metadata for every vehicle. And the European Court of Human Rights held in Catt v. United Kingdom (2019) that lawful collection does not license indefinite retention, and that the absence of a time limit was itself the violation.

What a European style framework would ask of Apex

1. A specific legal basis in statute, not a council resolution or a purchase order, naming the purpose and the places where scanning may occur.
2. A hard retention ceiling measured in days or months rather than years, with immediate deletion of non-matches, and images deleted within a day.
3. A published impact assessment completed before a single camera goes live, and kept current.
4. An independent oversight body and a code of practice the Town is legally obliged to heed, with mandatory logging and audit.
5. Enforceable individual rights of access, rectification and deletion, with a named point of contact and a regulator that can order the system switched off.
6. Necessity and proportionality tested against the actual local problem, with access tiered by data age and offense seriousness.
7. Signage and public notice at the point of collection, before enforcement begins.
8. No revenue interest and no vendor compensation that scales with reads, alerts or citations.

10Why this is not like your phone

The most common reply to any plate reader objection is that everyone already carries a tracking device voluntarily. It is a serious argument and it deserves a specific answer rather than a rhetorical one. There are six differences, and every one of them runs the same direction.

One: consent

• You choose to buy a phone. You accept terms. You can deny location permission, turn location services off, delete the app, use airplane mode, or leave the phone at home. Every one of those is a setting, and settings are the mechanism of consent.
• A license plate is issued by the state, must by law be displayed, cannot be removed, and is read whether you agree or not. There is no setting. The only way to opt out of a plate reader network is not to drive, in a town built around driving.
• Nobody in Apex was asked. The program was approved on a consent agenda in January 2025 with no public hearing, and nothing in North Carolina law required one.

Two: who gets captured

• Phone location data exists inside a customer relationship. Someone chose the carrier, chose the operating system, chose the apps.
• Plate reads capture every vehicle that passes, including people who are not customers, not suspects, not residents and not aware the camera exists. In LAPD’s holdings, 99.9 percent of 320 million stored plate images were of vehicles on no hotlist at all. The overwhelming majority of what a plate reader network holds is a record of innocent people going about their day. That is not an unfortunate side effect. It is the design.

Three: the legal process required to get it

• After Carpenter, obtaining historical cell site location from a phone company generally requires a warrant supported by probable cause. After Chatrie, a government request to Google for location history is a search regardless of how little history is sought.
• In North Carolina, an officer can query up to 90 days of a resident’s plate history with no warrant, no case number requirement, no logged justification and no independent audit. The identical category of information, a person’s movements over time, is protected by a warrant requirement when it comes from a phone and by an internal policy when it comes from a camera.
• That asymmetry is not a considered legislative judgment. It is a gap, and it is the gap every documented misuse pattern sits in: Denver’s 17 percent undocumented searches, the case number field defeated by typing “hehehe,” the Brunswick County resignation, and the California audit finding that nobody had audited anything.

Four: aggregation, which is the real subject

• One plate read is trivial and the vendors are right about that. It is a photograph of a car on a road.
• The Norfolk litigation put numbers on what a modest network produces. From 176 cameras, one plaintiff was logged 526 times in about four and a half months. Another was logged 849 times. That is four to six recorded sightings per day, per person.
• That volume is not a set of car photographs. It is a pattern of life: which mornings you leave late, which church, which clinic, which school, whose driveway on a Thursday evening, which meeting, and the time of day for each. As the Massachusetts Supreme Judicial Court put it, the whole of a person’s movements over a month “is not actually exposed to the public because the likelihood a stranger would observe all those movements is not just remote, it is essentially nil.”

Five: what deletion actually means

• This is where the popular version of the argument overshoots and the accurate version is still damning. It is not that plate data can never be deleted. It is that the Town cannot verify deletion, and that copies leave.
• Retention is a vendor side setting. Apex’s is 30 days. Flock’s new recommended default is 7 days. An LAPD audit found a Flock contract permitting five years. The number is a configuration choice made in a system the Town does not operate.
• Search results leave the platform. Exports, screenshots, PDFs and case files follow their own retention schedules, at the Apex Police Department and at every other agency that searched. A read that has been searched, downloaded and filed into a case in another state is now governed by that state’s records law, not by Apex’s 30 day setting.
• And the audience is large. Roughly 7,000 agencies and organizations can reach a participating department’s data. There is no delete button that reaches into another agency’s case file, and there is no mechanism by which the Town of Apex can recall a record it has already shared.
• Deletion is also an unobservable event. When Flock reports that data was deleted, the Town is taking a vendor’s word for something happening inside the vendor’s own storage environment. The draft governance ordinance circulating locally is right about the remedy: an independent third party audit, funded by the Town, conducted by an auditor with no financial relationship to any ALPR vendor, verifying deletion at the infrastructure level. Without that, retention policy is a promise rather than a control, and Denmark demonstrated what promises are worth.
• The precise sentence to use, rather than the slogan: plate data, once collected and shared, is outside the Town’s control, cannot be recalled, and the Town has no way to confirm it was destroyed. “Once it is on the internet it never leaves” is a slogan and it will be challenged. The sentence above is defensible and says the same thing.

Six: reciprocity, and who you can complain to

• Google and Apple can be sued by their own customers, are subject to consumer privacy statutes in a dozen states, answer to regulators, and answer to a market that can walk away. The relationship is asymmetric but it is a relationship.
• A resident photographed by a Town camera is not a customer. There is no contract, no terms of service, no account, and no ability to leave. And under Article 3D, North Carolina provides no private right of action for misuse, only a Class 1 misdemeanor that somebody else has to decide to charge.
• Which means the only accountability mechanism that exists for an Apex resident is this Council. That is not rhetoric. It is the actual structure of the law.

11Axon, the proposed alternative

The Public Safety Committee is expected to recommend dropping Flock and adding Axon. Section 12 compares the two contracts term by term. This section answers the prior question, which nobody at a public meeting has yet been asked to answer: what is Axon, what does it actually sell, who owns it, and what does buying from it commit the Town to. Everything below is sourced to Axon’s own filings and press releases, to federal court records, to SEC and Office of Government Ethics filings, or to named reporting, and the distinction between what is confirmed and what is alleged is marked in every case.

11.1 What Axon is

• Founded 1993 as AIR TASER, renamed from TASER International to Axon Enterprise in April 2017. Founder Patrick “Rick” Smith has been chief executive throughout. It is a public company, NASDAQ ticker AXON, which means its finances, ownership and governance are auditable by anyone. Flock is private and largely is not.
• FY2025 net sales $2.78 billion, up from $2.08 billion in 2024 and $1.56 billion in 2023. Second quarter 2026 revenue $904 million, up 35 percent year over year. 80,572,201 shares outstanding as of March 31, 2026.
• The recurring revenue figures are the procurement story, and they are Axon’s own. Annual recurring revenue $1.35 billion at the end of 2025, reaching $1.6 billion by mid 2026. Net revenue retention 126 percent, meaning the average existing customer spends 26 percent more each year than the year before, before any new customer is counted. Future contracted bookings $14.4 billion at the end of 2025, up 43 percent, which is roughly five times a full year of revenue already locked into signed multi year agreements.
• One number cuts against the growth story and is worth knowing. FY2025 GAAP net income was $124.7 million on an operating loss of $62.1 million, down from $377.0 million of net income in 2024. Axon is buying growth.

11.2 What Axon sells, confirmed against its own filings

The Town’s discussion has treated Axon as a plate reader vendor. It is not primarily that. The catalog below is drawn from Axon’s FY2025 annual report on Form 10-K, its investor releases and its product documentation.

• One historical fact that reframes the whole comparison. When Axon launched Fleet 3 in June 2021, it resold Flock Safety’s fixed plate readers, describing them in its own release as “Flock Safety’s industry-leading and ethically-designed fixed ALPR solution.” By 2025 Axon had built its own, and its FY2025 10-K now lists Flock Safety as a competitor, alongside Genetec, Motorola Solutions, Neology, PlateSmart and Rekor. The two companies the Town is being asked to choose between were commercial partners four years ago.

11.3 The consolidation question, which is the real objection

Axon now sells every link in the chain: 911 call intake, dispatch, the officer’s weapon, the officer’s camera, the car’s camera, the drone overhead, the fixed cameras on the poles, the aggregation layer that watches all of them at once, the cloud that stores the evidence, and the software that writes the report. No other single vendor covers that span. For a town, the question is not whether that is impressive. It is what happens to the Town’s leverage once one company holds all of it.

The clearest description of the mechanism is not from an advocacy group. It is from a federal antitrust class action brought by municipalities against Axon, In re Axon Enterprise, Inc. Vievu Antitrust Litigation, D.N.J. No. 3:23-cv-07182, led by the Mayor and City Council of Baltimore with the Township of Howell, New Jersey. Motions to dismiss were denied in part on January 31, 2025, so the body camera damages claims are live. The following are allegations sworn in federal court by other municipalities, not judicial findings, and should always be described that way.

• Axon’s own answer, quoted fairly. In its SEC filings Axon describes the case as “a purported antitrust class action brought by three municipalities based largely on the FTC’s unproven allegations.” That is a legitimate characterization and the Council should hear it alongside the allegations.
• The FTC case those allegations came from, and why it ended without an answer. On January 3, 2020 the Federal Trade Commission voted 5 to 0 to challenge Axon’s completed 2018 purchase of Vievu, its closest body camera competitor, alleging the deal “reduced competition in an already concentrated market,” that “competition between Axon and Vievu resulted in substantially lower prices,” and that side agreements barred Vievu’s former parent from competing against Axon products with some restrictions running beyond a decade. The FTC dismissed its own case on October 6, 2023, citing “the increasingly unlikely possibility of reaching a timely resolution” after years of delay. The merger was never adjudicated on its competitive merits.
• And the Supreme Court case is not the vindication it is sometimes described as. Axon Enterprise, Inc. v. FTC, 598 U.S. 175, decided April 14, 2023, held only that a federal district court may hear a constitutional challenge to an agency proceeding without waiting for that proceeding to finish. The Court did not rule that the FTC’s structure is unconstitutional and said nothing about the merger. That procedural win produced the delay the FTC then cited in giving up.
• Axon markets sole-sourcing to its customers, in writing. Its own page, “4 Reasons Why Agencies Have Opted to Sole Source,” argues that an agency is not “buying simply a camera” but “a sophisticated hardware-software ecosystem,” and that running a competitive solicitation after field trials is “overly bureaucratic and wasteful.” Axon is also available through the OMNIA Partners cooperative purchasing vehicle, which agencies use in place of their own competitive bidding.

11.4 Fusus, and why the camera count stops being the question

• Fusus is not a camera. It is an aggregation layer. In Axon’s own words it “excels in aggregating live video, data and sensor feeds from virtually any source” and “supports devices and sensors from dozens of different providers.”
• What it can pull in: the department’s own body and fleet cameras, pre-existing municipal cameras in traffic, parks, transit and schools, private and business cameras through a feature called Community Connect that lets residents and businesses opt in to share their feeds, drone video, and plate reader overlays, all on one live map with AI alerting through Axon Vision. Miami Beach has integrated 1,600 or more cameras this way.
• A related private-camera channel opened in October 2025. Axon and Ring announced a partnership letting officers send geofenced video requests from Axon Evidence to Ring doorbell owners through the Neighbors app, with shared clips flowing into Axon Evidence under chain of custody. Sharing is voluntary.
• So the surveillance question changes shape. It stops being how many cameras the Town is buying and becomes how many camera feeds the platform will be able to see, and who decides when a new one is added. A ten camera plate reader contract that includes Fusus can, in principle, become a live view into hundreds of existing public and private cameras with no further procurement and no further vote.
• One genuine point in Axon’s favor, stated fairly. Axon describes its plate reader sharing as invite-based: data stays with the originating agency by default, and an agency may “invite a specific partner agency to search matches from your cameras, on terms you control and revocable at any time.” That is architecturally different from a default national lookup network, and if true it is a real advantage over Flock’s historical model. It appears in Axon’s marketing and product pages. It was not located as a binding term in Axon’s published master agreement. Require it in writing or treat it as unbought.

11.5 Who actually owns Axon

Because Axon is publicly traded, this is answerable from primary documents rather than from rumor. The table below is taken from Schedule 13G ownership filings and from the chief executive’s most recent Form 4, against the 80,572,201 shares outstanding reported in Axon’s 2026 proxy statement.

• Roughly 80 percent of Axon is held by institutions, about 15 percent by retail investors and 4 to 5 percent by insiders. These are overwhelmingly index and pension funds that hold Axon because Axon is in the indexes. Their ownership carries no political implication of any kind. Some financial websites report institutional ownership above 90 percent; those figures come from summing 13F filings, which double-counts the same shares reported by both a custodian and a manager, and should not be quoted.
• No single holder is anywhere near control. The top seventeen shareholders together hold roughly half the company, and no Schedule 13D has ever been filed on Axon Enterprise by anyone, which is the filing that would disclose an activist or control stake.
• Rick Smith is a consistent net seller under a Rule 10b5-1 trading plan adopted May 12, 2025. His Form 4 figure of about 3.8 percent excludes exercisable options and vested restricted stock, so the beneficial ownership figure in Axon’s proxy is higher; read pages 29 to 31 of the 2026 proxy for that number before quoting one.
• The board is eleven directors, ten of them independent, chaired by Michael Garnreiter, with Rick Smith the only insider. Jeri Williams, the former Chief of Police of Phoenix, sits on it, which is worth knowing when Axon’s public safety credentials are asserted. No Trump-affiliated individual sits on Axon’s board.

11.6 The Trump question, answered precisely in both directions

A claim circulating locally holds that Axon’s investors include Donald Trump and his family. Half of that is documented from a primary government filing and half of it is false, and the half that is true is not an ownership claim. Getting this exactly right matters, because it takes ninety seconds to check and a presentation that overstates it will lose the room.

• What is false: that Trump, his family or Trump entities are investors in Axon in any ownership sense. The arithmetic settles it. One to five million dollars against roughly 80.6 million shares trading in the hundreds of dollars is about 0.005 to 0.01 percent of the company. That is hundreds of times below the five percent threshold that triggers an SEC ownership filing, which is why a search of Axon’s SEC filings finds nothing: there is no Trump Schedule 13D or 13G on Axon and there mathematically could not be one at that size. No stake in any ordinary sense, no board seat, no control, no influence over the product.
• What is false: that Donald Trump Jr. is connected to Axon. He holds roles at 1789 Capital (partner), Unusual Machines (advisory board, November 2024), Dominari Holdings (board of advisors, February 2025, with Eric Trump), PSQ Holdings / PublicSquare (board, December 2024) and GrabAGun (board), among others. A watchdog group notes he joined one company board between 2020 and the 2024 election and roughly ten afterward. Axon is on none of those lists, and 1789 Capital’s published portfolio does not include Axon.
• Where the confusion almost certainly comes from, and it is an easy mistake. Unusual Machines is a drone company, and Axon is now also a drone company through Dedrone and Axon Air. “Trump family, drones, police technology” collapses into “the Trumps own the Taser company” with almost no friction. On the firearms side the equivalent is GrabAGun. Separately, Axonius and Axonics are unrelated companies with confusingly similar names.
• No Trump Organization or Trump Media stake in Axon appears in any filing or credible report. The word “Trump” appears in Axon’s own annual report only in a tariff risk factor, not as a shareholder, investor, partner or customer.

11.7 The documented political relationships, which matter more than the stock

• Federal lobbying. Axon spent $510,000 in the first quarter of 2025 and $550,000 in the second, against roughly $1.5 million across all of 2024. Registered subjects include body-worn cameras, less-lethal technologies, digital evidence management, drone policy and the FY2026 defense authorization. At least eight outside firms were retained in 2025 alongside its in-house team, one of them registering specifically for “TASER, counter-drone, de-escalation drone policy.” OpenSecrets reports that Palantir, Axon and Parsons together raised combined lobbying spending 194 percent from 2020 to 2025 while winning ICE contracts.
• A corporate PAC. The Axon Enterprise Inc Political Action Committee, FEC committee C00890590, was registered October 2, 2024. For the 2025 to 2026 cycle it reports $135,000.01 in receipts, all from individual contributions, and $101,000 in disbursements of which $100,000 went to other committees. Its recipients and any party split could not be retrieved from FEC data for this briefing and no claim about them should be made.
• The revolving door, which is the strongest item in this section. Ronald D. Vitiello served as Trump’s Acting Director of Immigration and Customs Enforcement from June 2018 to April 2019; the President withdrew his nomination for the permanent post saying he wanted to go “in a tougher direction.” He then spent more than five years at Axon as head of Department of Homeland Security programs and strategy, under a press release Axon titled “Axon Welcomes Former ICE Director Ronald D. Vitiello as Head of DHS Program and Strategy.” He returned to government in January 2025 as a senior adviser at Customs and Border Protection and was named Acting Deputy Commissioner of CBP in 2026, while Axon pursues DHS and ICE business. An Axon lobbyist also served as a transition adviser to the Secretary of Veterans Affairs.
• Federal contracts. ICE bought roughly $5.1 million of body cameras and $22,376 of Tasers in March 2025; ICE Enforcement and Removal Operations bought $2.6 million of Tasers in 2020 and 2021; a DHS body camera contract has been reported above $30 million; and the February 2026 solicitation above is for about $220 million.
• Read that paragraph against the reason this debate started. Apex’s own controversy is about two federal agencies reaching its camera data. If federal access is the concern, Axon’s federal relationship is larger, older, formally contracted and staffed by former federal enforcement leadership. A switch made to solve federal access does not obviously solve it.

11.8 The governance record, which is the fair criticism

• Axon’s own ethics board walked out. On June 6, 2022, nine members of the AI Ethics Board Axon had convened in 2018 resigned over the Taser drone proposal. This was not a decorative body; it was staffed and supported by the Policing Project at New York University School of Law, whose director chaired it, and it produced substantive published work. Their statement records that “a majority of this Board, by an 8 to 4 vote, recommended that Axon not proceed,” that “for years the Board has warned Axon against the use of real-time, persistent surveillance,” and that the announcement “bypassed Axon’s commitment to consult” them. Axon’s own hand-picked, academically staffed ethics body resigned over a product proposal.
• That board’s plate reader report is the single best source of questions for the meeting. It published a dedicated report on automated plate readers with fifteen recommendations, warning about false positives, long-term tracking of innocent drivers, and disproportionate enforcement burdens on low income communities and communities of color, grouped into democratic regulation, vendor design self-restraint, and limiting enforcement to serious offenses. Asking Axon’s representative, on the record, which of Axon’s own former ethics board’s fifteen plate reader recommendations the proposed contract satisfies is precise, fair, and very hard to deflect.
• That board’s earlier face recognition report concluded the technology was “not yet reliable enough to justify its use on body-worn cameras,” citing performance disparities across racial, ethnic and gender groups. Axon accepted the recommendations and committed in 2019 not to build face matching products. Whether the company still considers itself bound by a commitment made to a board that then resigned is not something this briefing can establish. Ask in writing.
• Draft One is the most concerning single fact about Axon’s auditability culture. Axon’s AI report writer does not retain the original AI-generated draft, so after the fact it is impossible to determine which parts of a police report were written by software. The Electronic Frontier Foundation, in July 2025, quotes an Axon senior product manager explaining the design as avoiding “disclosure headaches,” and reports that audit logs record only that Draft One was invoked, not which text it produced. The Palm Beach County Sheriff’s Office was able to identify more than 3,000 AI-generated reports from December 2024 to March 2025 only by keyword-searching for the disclosure language. Axon’s product page advertises Draft One as auditable. Both statements can be literally true, which is exactly why any vendor’s claim to be “auditable” must be answered with a specification of what precisely is logged and for how long.
• And the absence of Axon plate reader scandals is not a clean bill of health. The fixed line launched in April 2025. Flock’s problems are documented because Flock has been deployed at scale since 2017. A council should read “no reported problems” as unknown, and weigh the adjacent record instead: the ethics board walkout, the Draft One findings, and two separate competition proceedings brought by the federal government and by other municipalities.

11.9 What happened when other departments actually used Axon’s AI

Axon’s AI products are the fastest growing part of its business and the part with the most independent evidence against them. In July 2026 Forbes published an investigation by Thomas Brewster built on police emails and AI-generated reports obtained through public records requests in several jurisdictions. It is the most useful document available on whether Axon’s AI does what Axon says it does, and unlike most technology reporting it rests on the departments’ own words rather than on the vendor’s.

• The scale, and the price. Axon says Draft One is in use by 600 police departments, has helped write 600,000 police reports and has saved an estimated 300,000 hours, the last figure drawn from Axon’s own customer surveys. Access comes through the AI Era Plan at $199 per officer per month. Axon booked roughly $750 million in AI Era contracts in the year after launch, and reported AI revenue growing more than 700 percent year over year in the first quarter of 2026, without releasing actual sales figures. Its chief executive told shareholders that “AI is the greatest accelerant Axon has ever had.”
• The only peer-reviewed test of the central claim found no time savings at all. Ian Adams, a criminology professor at the University of South Carolina and a former police officer, studied the Manchester, New Hampshire police department’s use of Draft One and published the results in the Journal of Experimental Criminology in 2024. He compared timestamps for when officers started and filed reports, with only some officers having access to the software. There was no improvement, because the time saved on drafting was consumed by editing: removing irrelevant material, correcting inaccuracies and adding facts the model had missed. A Manchester lieutenant who took part put it plainly to Forbes: “It was just easier to type the report themselves.”
• The finding a council should sit with longest is about testimony, not technology. Despite the timestamp data, about half of Manchester’s officers believed Draft One had sped up their report writing. Adams’s conclusion: “There’s a definite gap between the perception and objective reality.” That single sentence is the reason officer testimonials, including the ones a vendor will bring to a public meeting, are not evidence of anything. It applies equally to Flock and to every enthusiastic quote in Section 3 of this briefing.
• Two departments have already quit. Manchester, the first agency in the country to test Draft One, dropped it in 2024. The Anchorage Police Department in Alaska also stopped using it in 2024, citing zero time savings.

• Axon’s response, given in full because it is partly reasonable. A spokesperson said Form One is in early access, a stage intended for agencies to give feedback before wider release, and that officers retain full control over the final report. A Lafayette sergeant confirmed the emails came from a testing period in which officers had been explicitly asked what could be improved, and said several of the issues raised were then addressed, though he did not answer follow-up questions about what specifically was fixed, and the department has not decided whether to keep the tool. On the Manchester study, a second Axon spokesperson called it “an interesting historical snapshot” whose conclusions “rely on AI-assisted narratives generated in 2024 from a single agency,” and said the product has “evolved significantly in the last two years.”
• But that defense has a hole, and Forbes names it. There are no recent peer-reviewed studies testing whether the technology has improved at writing police reports. The vendor’s position is that the only independent evidence is out of date, and no newer independent evidence exists. For a council, that is not a reassurance. It is a description of buying on faith, and it is the same epistemic position this briefing has documented on plate reader accuracy in Section 3.

• The retention finding is the one that should trouble this Council most, because it is the same argument as the cameras. Forbes obtained a cache of AI-generated reports from the Richmond, California police department. They included callouts to schools over bullying of pupils and teachers where no crime occurred, deaths by natural causes, and graphic descriptions of sexual assaults. In one, a teenage student typed a line from the cartoon SpongeBob SquarePants on a laptop, a classmate misheard it as a threat, staff called the police, and the officer established within minutes that it was a misunderstanding. The report nonetheless recorded that the incident would be “retained for future pattern tracking.” A minor, no crime, no charge, and a permanent entry in a pattern-matching system. That is the plate reader retention argument in Section 10, transplanted into a child’s school record, and it arrived as a side effect of a product bought to save time on paperwork.
• The civil liberties framing, from a named specialist. The senior counsel who directs the AI and Human Rights Program at the Electronic Privacy Information Center told Forbes that the practice is “making what was supposed to be an accountability mechanism another tool of public surveillance.” The body camera was introduced to hold police accountable. Feeding its audio into a report generator, and its video into the aggregation layer described in 11.4, turns the accountability device into a collection device.
• One critical voice to discount, and one to credit. The chief executive of a competing police technology firm told Forbes that AI drafts are “worse than starting with a blank page, because now you’ve given police this corrupted report,” and dismissed Axon’s product in considerably blunter terms. He sells a rival tool, so treat the vendor-on-vendor attack as interested, while noting that the mechanism he describes is the same one the peer-reviewed study measured. Against that, the chief of the Avon, Colorado police department told Forbes that a suite of AI tools genuinely eased the administrative burden, describing officers who would otherwise face “five or six reports” at the end of a shift and calling the benefit one with “a very beneficial mental health aspect.” That is a real benefit, and a timestamp study does not measure it. Both things are true, which is why the answer is a pilot with published metrics rather than a purchase.

12Flock versus Axon, term by term

Section 11 described what Axon is. This section is about whether switching to it solves the problem that started this debate. The short answer, on the specific concerns that drove it, is that on retention, federal access and auditability, Flock’s published commitments as of August 2026 are more concrete and more restrictive than anything located in Axon’s public contract. That is not a defense of Flock. It is a warning that a switch made for those reasons may not deliver them.

The terms, side by side

On Flock’s August 2026 changes, which are material to any comparison and which the ACLU characterizes as “a thinly veiled PR attempt to counter communities’ genuine privacy concerns”: default retention cut from 30 days to 7; data sharing off by default; offense filtering so a community can permit searches for stolen vehicles, missing persons or violent crime while blocking immigration related searches; mandatory audit assistance and mandatory case codes by the end of 2026; automatic suspension of users flagged for abnormal activity; mandatory multi factor authentication; removal of Flock’s own discretion to disclose data to government agencies on imminent harm grounds; removal of restrictions on municipal non-appropriation; and an independent security review published in September 2026.

The practical implication for Apex is worth stating plainly: Flock’s new defaults are now stricter than Apex’s negotiated settings. The Town is at 30 days with sharing that admitted two federal agencies. The vendor’s own recommendation is 7 days with sharing off. That gap is a decision the Town has not yet made, rather than a limitation of the product.

Why the swap may not solve the problem

• Denver is the template, and it proves both halves of the point. Denver dropped Flock and advanced an Axon contract in March 2026 with 50 cameras instead of 100, 21 day retention, no access to a vendor operated national database, and mandatory audit trails for every query. Every one of those four terms is absent from Axon’s standard master agreement. They had to be negotiated. That proves Axon will agree to them, and it proves a town will not get them by default.
• Syracuse is the cautionary tale. Its council approved an Axon plate reader contract in February 2026 and then voted unanimously in March 2026 to revoke Flock’s access over federal data sharing. Axon supplies body cameras, accessories, licenses and software to Department of Homeland Security components, confirmed on the record by a DHS spokesperson. Flock’s spokesperson made the obvious self interested point, and was factually right: “Claiming different data is collected, then going to Axon is crazy,” and “Then you can’t install Axon. Then you just have to not use LPRs.” That second sentence deserves an honest answer rather than a dismissal.
• If the concern is federal access, a Flock to Axon switch does not on its face address it. Axon’s federal relationship is larger, older and formally contracted, and it operates FedRAMP High federal facing cloud infrastructure. Note carefully what that does and does not mean: FedRAMP authorization does not give federal agencies the ability to read local agency data. It means Axon runs federal facing infrastructure and has a large and growing federal customer base including DHS.
• If the concern is vendor dependence, the switch concentrates it. Axon likely already holds the Town’s body cameras and its evidence system. Add 126 percent net revenue retention, a $15.1 billion contracted backlog, no termination for convenience, and plate readers designed to feed Fusus which feeds evidence and records. Leaving Flock for Axon does not reduce vendor dependence for most departments. It deepens it.
• If the concern is auditability, Axon is the weaker documented vendor today. Flock now has published, dated, mandatory commitments. Axon has a trust center checkbox and marketing references, no audit log clause in its master agreement, and Draft One as a documented instance of designing away an audit trail.
• The procurement risk is specific and it is the one raised in local discussion. If the Town already holds an Axon master agreement for body cameras, adding plate readers by amendment or by quote under that agreement means no competitive solicitation, no fresh public hearing, and a new surveillance capability inheriting terms negotiated for a different purpose. Under Axon’s structure, subscription terms are set in the quote rather than the master agreement. A surveillance system should not arrive as a line item on a quote under a body camera contract. Retention limits, immigration search prohibitions, audit log access and the program opt out are not in the standard agreement and would each have to be negotiated into an ALPR specific addendum.
• And the absence of Axon plate reader scandals is not a clean bill of health. The line launched in April 2025. Flock’s problems are documented because Flock has been deployed at scale since 2017. A council should read “no reported problems” as unknown, and weigh Axon’s adjacent record instead: the ethics board walkout, the Draft One findings, and two separate competition complaints.

13Four options, and what each requires

Mayor Gilbert framed three: continue with Flock, change providers, or discontinue. There is a fourth, and it is the one Flock’s own August 2026 changes made available.

Option 1: renew Flock as currently configured

• Requires no action and answers none of the questions in this briefing.
• It also leaves the Town at 30 day retention when the vendor now recommends 7, and with a sharing configuration that admitted two federal agencies without anyone at the Town noticing.

Option 2: switch to Axon

• Requires negotiating an ALPR specific addendum, because the four terms Denver obtained (camera cap, stated retention, no vendor operated national database, per query audit trails) are not in the standard master agreement, along with a written opt out of the customer data program the Town would otherwise be enrolled in by default, and a termination clause the standard agreement does not contain.
• Does not address federal access, since Axon holds active DHS contracts, and concentrates vendor dependence in the company that likely already holds the Town’s body cameras and evidence system.
• Watch the procurement route. If this arrives as an amendment or a quote under an existing master agreement, it will bypass competitive solicitation and a fresh hearing. Insist it come to Council on the record as a new acquisition.
• And ask about Fusus separately from the cameras, because the camera count and the platform’s reach are two different questions.

Option 3: discontinue

• The honest accounting requires naming what remains. State cameras remain: the SBI program on NCDOT rights of way was made permanent in July 2026 and covers roughly 80 percent of North Carolina roads, with a stated goal of cameras at every entrance to and exit from the state, and the Town has no vote over it.
• Private cameras remain, in HOAs, retail parking lots and hospital campuses, and the Town’s authority over them is limited to installer licensing (state occupied) and right of way encroachment (a real but partial lever).
• Neighboring towns’ cameras remain. Holly Springs, Garner, Fuquay-Varina, Knightdale, Wake Forest, Raleigh and Cary all operate readers, and Apex residents drive past them daily.
• Discontinuing is a defensible decision about what the Town itself will operate and pay for, and about whether Apex data feeds a national network. It is not a decision that makes Apex unsurveilled, and claiming otherwise invites an easy rebuttal.

Option 4: no ALPR without a governance ordinance

• Decide the configuration on the record rather than inheriting it, and adopt the governance framework first, so that whichever vendor the Town uses, and whenever the law changes, the rules belong to Apex rather than to a vendor’s default settings.
• This is available precisely because Flock’s own new defaults are stricter than the Town’s current negotiated settings, so tightening materially does not even require changing vendors.

The seven terms that carry the weight

Drawn from the specific failures documented in this briefing rather than from a template, and applicable under any of the options above.

1. Retention at or below the state floor, with verified deletion. A short number plus an independent verification that deletion actually occurred at the vendor’s storage environment, because Denmark showed what a retention limit is worth without an auditor, and Littleton showed that deactivation is not self executing.
2. No participation in national or federated lookup networks. Case by case sharing only, with supervisory approval, and federal access by written request or warrant only. This is the specific failure Apex already experienced.
3. Per query logging of case number, purpose and factual basis, validated against the records management system, with a sample of searches reviewed against the underlying case files. An unvalidated free text field is defeated by typing “investigation,” and has been.
4. An independent annual audit by an auditor with no financial relationship to any ALPR vendor, because self audit produced no findings in every case examined for this briefing, including two where the answer to “when was the last audit” was “never.”
5. Mandatory independent visual verification before any stop based on an alert. This single requirement addresses the Gilliam, Green, Morristown and Sherwood fact patterns directly, and Virginia chose it in statute instead of a warrant requirement.
6. Council approval for any new capability, defined to include new analytic products and new data sources applied to existing collection, with $0 pilots, trials, grants and donations treated as acquisitions, and with amendments, renewals and extensions requiring approval. San Diego and Denver are the two cases showing why each clause is needed.
7. An automatic consequence for audit failure that does not require a discretionary decision, such as suspension of the program pending Council review, rather than “the Chief may review.”

14Executive summary

• Apex bought network membership, not ten cameras. The $71,750 contract, approved 5 to 0 on a consent agenda in January 2025 with no public hearing, enrolled the Town in a network of roughly 120,000 cameras that about three quarters of Flock’s police customers can search, and which reportedly makes a participating department’s data reachable by around 7,000 agencies. This renewal is the Town’s first real deliberation on it.
• Flock was built to sell cameras to HOA boards, which were more than 40 percent of its business into 2021, and the consent model was set there: whoever buys the camera decides who searches the data, and the people photographed are never asked.
• The technology genuinely works at one thing: rapidly locating a specific vehicle already on a list. The rescues in Section 3 are real, several almost certainly saved lives, and one of the strongest is a Wake County child abduction case from July 2026.
• It does not work at the things it is sold on. Randomized trials find no deterrent effect. The best clearance study found improvements that were not statistically significant. In the country’s largest deployment, clearance rates are flat or falling, with homicide clearance down from 53.4 to 48.0 percent. The strongest supporting study, unpublished, finds an 11 percent drop in vehicle theft and nothing about violent crime.
• No honest percentage of crime solved by plate readers exists. Flock’s own figures are, in its words, “directional estimates rather than audited totals”; its named academic validator said the underlying police data is “too varied and incomplete” for meaningful analysis; Forbes found two of its city crime claims wrong, one conceded by the benefiting police chief; and Virginia’s state crime commission concluded that “most of the information on ALPR successes, errors, and misuses is anecdotal.” The only publicly audited chain, in Austin, is one arrest per 455,000 plate reads.
• The harms are documented and paid for. $1.9 million in Aurora for four handcuffed children after a Colorado plate matched a Montana motorcycle. $495,000 in San Francisco for a misread digit. Thirty one documented wrongful stops with guns drawn in nearly two thirds, including two in North Carolina. Two thirds of the errors are human, not optical, so a better camera fixes almost none of it. Alert level error rates on the record: 32 percent in Los Angeles, 37 percent in Vallejo, 71 percent in Roseville.
• The worst documented harm is misuse, not error. The national abuse database listed 197 incidents on September 3, 2026. Tabulated in Section 5: 75 percent is deliberate human misuse and only 15 percent is technical error. One in four records produced an arrest or indictment, 51 people lost their jobs, and the largest single case is an officer with more than 10,000 searches facing 100 felony counts. Georgia accounts for 68 records because Georgia audited, which means the table measures auditing rather than misconduct, and Apex has never audited its own searches.
• Warrantless use is currently lawful and the law is moving. Every ALPR ruling has gone the government’s way on the facts, but the leading one is on appeal in the Fourth Circuit, which governs North Carolina, was decided five months before a Supreme Court decision that rejected duration thresholds, and says on its face that the balance “could conceivably tip the other way” as camera numbers grow. Any contract signed now should be terminable without penalty.
• Nothing in North Carolina law gives this Council a vote. No notice, no hearing, no referendum, and because ALPR is a service subscription, the competitive bidding statutes do not reach it at any dollar figure. The state caps retention at 90 days and regulates querying not at all: no warrant, no case number, no logged justification, no independent audit. Whether Apex ever votes on surveillance technology depends on whether this Council writes that requirement itself.
• The Town holds the financial risk. Flock’s total contractual liability is roughly $35,875, it disclaims any warranty as to results, and its indemnity covers pole installation rather than misread plates. Apex gets no qualified immunity, federal civil rights exposure sits outside the state insurance immunity cap, fee shifting means a nominal verdict still costs six figures, and the Town has no published answer to whether it carries the optional law enforcement liability and cyber coverage lines.
• The 2024 utility billing attack establishes one narrow, hard to rebut point. 22,601 people exposed, including Social Security and passport numbers, 21 months to notify, 19 months of billing disruption, $5.9 million in delayed revenue against $381,000 in arrears before the attack, and total cost still undisclosed. A billing failure could be rebuilt. A plate database failure is a disclosure, and there is no backup to restore.
• Both comparisons cut against the current configuration. North Carolina requires an act of the General Assembly before a town may photograph a car running a red light, and that narrow tool still lost public consent in nine state bans and a dozen referenda. Europe permits plate readers only with a statutory basis, retention measured in days to months, a published impact assessment, tiered access, mandatory audit and individual rights of deletion. Germany permits the scan and forbids the keeping.
• Switching vendors does not answer the objections. On retention, federal access and auditability, Axon’s published terms are weaker than Flock’s current ones, its real risk is the Fusus aggregation layer rather than the cameras, Denver had to negotiate the four terms that matter because none come standard, and Syracuse voted out one vendor over federal data sharing and voted in one with active DHS contracts.
• The claim that Trump or his family are Axon investors is false in the ownership sense and should not be used. The true and narrower version, a documented $1 million to $5 million personal purchase two weeks before a matching $220 million ICE solicitation, is a federal conflict of interest question, not a fact about who owns the company.
• The decision is about configuration, not about whether cameras ever help. Retention length, camera density, sharing scope, query controls and who approves the next capability are all Council choices, and they are the same variables the vendors themselves cite as the reason the technology is lawful.

15The national argument

In August and September 2026 the most consequential actions against plate readers came from Republican officials, and the strongest operational objections came from police chiefs. Any presentation that frames this as a partisan issue is describing a debate that no longer exists.

Florida

• August 28, 2026: Governor Ron DeSantis directed action. August 31, 2026: a one page Florida Department of Transportation memo revoked all permits for automated license plate readers in state highway rights of way, giving agencies 30 days to remove the devices. Existing permits were “hereby revoked.” It does not reach cameras on municipal streets or private property, and DeSantis acknowledged the limit: “The only way it could be dealt with is legislatively.”
• DeSantis, at a news conference the same week: “If you want to get in your car and go down to Buc-ee’s in St. Augustine, it’s really none of the government’s damn business if you’re doing that, right?”
• “What we don’t want in the state of Florida is to have a digital AI surveillance state where everything we’re doing is tracked at all times.”
• “Where is this data going? They claim companies won’t misuse it, but no one’s watching that.”
• “I guarantee you every sheriff in Florida can point to examples where it’s been used appropriately and effectively. But you also have to say, how could this be misused?” That is the fairest one sentence statement of the problem anyone has offered.
• Five Florida sheriffs dropped plate readers the same week: Liberty County (Sheriff Dusty Arnold), Clay County (Sheriff Michelle Cook), Duval County and Jacksonville (Sheriff T.K. Waters), Putnam County (Sheriff H.D. DeLoach, who disconnected 18 readers immediately), and Pasco County (Sheriff Chris Nocco, terminating by September 30). Sheriff Arnold cited “the continuing constitutional and privacy concerns surrounding license plate reader system.”
• Sheriff Waters complied while still calling plate readers “an enormously valuable investigative tool.” That is the honest separation of a tool’s value from its governance risk, and it is a useful model for how a police chief can support the technology and still accept a limit.
• Florida’s legislature has not acted. A bill barring AI companies from selling personal information passed the Senate 37 to 2 and the House declined to take it up, so Florida’s action so far is executive.

Texas

• September 1, 2026: Governor Greg Abbott ordered state agencies to stop funding Flock cameras. “Citizens are concerned about their Fourth Amendment rights. They have to be protected. Their privacy must be protected.” His spokesman noted that “almost all Flock cameras in Texas are used by cities, and those cities are increasingly canceling their contracts.”
• Wylie and Pflugerville ended their contracts; Princeton is reconsidering; Liberty Hill is deciding by the end of October; and the mayor of Manvel publicly reversed his support.
• Hood County Constable John Shirley covered the cameras himself and the commissioners court voted to disconnect them: “permanent, warrantless tracking doesn’t belong in a free society.”

The United States Senate

• August 26, 2026: Senator Josh Hawley, chairman of the Senate Judiciary Subcommittee on Crime and Counterterrorism, opened a formal investigation by letter to Flock’s chief executive, with a document deadline of September 8, 2026. He demanded law enforcement access policies, camera placement records, data handling procedures, a list of suspected misuse incidents since 2021, accuracy and security details, and government contracts and investor information.
• “Americans do not surrender their privacy rights when they drive to work, drop their kids off at school, or go to church.”
• “Congress never authorized the network your industry has built. Your company’s internal policies are, in practice, the only safeguards for hundreds of millions of Americans.”
• “The overwhelming majority of the Americans captured in those records did nothing wrong.”

State legislatures, and the party split

• Pennsylvania has five pending proposals and four of them are Republican. A Senate bill would ban plate readers connected to networks that automatically identify and store vehicle information. Two House bills would impose a warrant requirement with retention limits and mandatory signage, and authorize local moratoriums. A second Senate bill would prohibit “mass, suspicionless collection” with carve outs for stolen vehicles, Amber Alerts, missing persons and warrants. Two Democratic bills would impose usage and storage limits, transparency and civil penalties.
• Arizona: the Attorney General opened a review of plate readers and Arizonans’ privacy rights in August 2026. A separate bill would require supermajority voter approval before a city deploys surveillance technology.
• Enacted restrictions in 2025 and 2026 include Virginia’s comprehensive statute, plus new laws in Idaho, Kentucky, New Mexico and Washington, and existing Illinois restrictions on immigration related sharing.
• North Carolina moved the other way, making the SBI’s highway program permanent through the July 2026 budget. A briefing implying the national trend is uniformly restrictive would be wrong, and the divergence arguably raises the stakes of a local decision rather than lowering them.

Civil liberties organizations, across the spectrum

• The Electronic Frontier Foundation’s formal position as of August 2026 is elimination rather than reform: “there is no configuration of an ALPR network that eliminates this risk, because the risk is the mass surveillance itself.” Its interim asks are warrant requirements, deletion deadlines, use and sharing restrictions, and bans on sharing with federal and out of state agencies.
• On the vendor switch specifically, EFF’s Dave Maass: “You’re still reliant on a third party company. If that data is still there, it can be abused.”
• The ACLU’s core objection is suspicionless collection. Jay Stanley: “It collects data on passing cars without individualized suspicion of wrongdoing.” The ACLU called Flock’s August 2026 changes “a thinly veiled PR attempt.”
• The Institute for Justice, a libertarian public interest law firm, maintains the abuse and error databases cited throughout this briefing and is litigating the Norfolk case. The same technology is being fought by a libertarian firm in federal court and by the ACLU in state legislatures.

Police and sheriffs who declined, which is the most persuasive category

• Los Angeles Police Department did not renew its three year agreement, which expired July 11, 2026, after its Inspector General documented a 32.3 percent error rate. Chief Information Officer Dean Gialamas: “This contract is not being renewed because of serious concerns around civil liberties and civil rights issues, particularly around privacy and the data that is being collected from these cameras.” The Inspector General recommended suspending new camera deployment and new contracts pending public input.
• Pflugerville, Texas. Police Chief Willie Richards declined to recommend renewal: “Clarity around how we protect community data is essential, and currently I do not have that clarity.” The council voted unanimously to terminate an $87,500 per year contract and disable 28 cameras.
• Willard, Missouri, August 2026. A chief who personally believes the cameras help solve crimes still declined to recommend renewal: “My recommendation that the board approve another year of funding for the Flock system was not supported by the data that I received and can verify from my own research.” He added that the public at large is against the system.
• Farmington, Arkansas. The police department itself stated that termination of the Flock contract should be considered.

The tally, stated honestly in both directions

• Do not present a single number as authoritative. The counts differ by a factor of two depending on whether deactivations, rejections and expirations are included.
• And the other side of the ledger is real. Flock says new city partnerships have outpaced non-renewals by roughly ten to one in 2026, and roughly 800 United States cities approved new contracts in the same period, against a base of around 120,000 cameras. The correct statement is that cancellations are accelerating sharply and the installed base is still growing. Both halves are true, and saying only the first invites correction.
• Cameras have also been vandalized in at least 36 states, which is worth naming for a different reason: it is evidence of exactly the loss of public consent that the United Kingdom’s own regulator named as the greatest risk to the technology.

Public opinion, including a correction

• The Fox News polling figures circulating locally could not be verified and appear not to exist. Fox News’s own official poll archive contains no poll on license plate readers, ALPR or government surveillance, and the specific figures in circulation, Republican support in the teens and Democratic support around 11 percent, are inconsistent with every poll that does exist and show a party pattern no poll reproduces. Do not use them. Replace them with the two below, which are more defensible and say the same thing.
• Knox County, Tennessee, August 15 to 17, 2026. Spry Strategies, 350 likely general election voters: 58.3 percent opposed and 37.5 percent supported police use of Flock cameras.
• Arizona statewide, August 17 to 19, 2026. NextGen Polling, 1,627 registered voters: 63 percent opposed plate readers, 82 percent concerned about data misuse, and 60 percent favor limiting their use.
• The most interesting finding is academic rather than political. A randomized experiment published in 2019 (Merola, Lum and Murphy, 405 respondents) found that explaining plate reader capabilities significantly decreased trust in police among the informed participants. Support for this technology declines as understanding of it rises, which is an argument for public deliberation and also an explanation of why programs so often begin on consent agendas.

The vendor’s answer, stated fairly

• Flock’s chief executive: “When I think about legislation that calls for an outright ban, that’s like banning vehicles,” and “Tools this powerful must come with guardrails this strong.”
• A Flock spokesman: “Rather than banning a critical public safety tool, we support thoughtful legislation and strong privacy, transparency, and accountability guardrails.”
• Flock’s Florida statement: the technology “is an important public safety tool for law enforcement agencies across Florida, helping officers solve serious crimes, find missing people, recover stolen vehicles and support safer communities.”
• That position is not unreasonable, and it concedes the entire premise of this briefing. If tools this powerful must come with guardrails this strong, then the guardrails are the decision in front of this Council, and North Carolina has not written them. Somebody has to, and the only body that can do it for Apex is sitting in this room.

16A field guide for Apex residents

Everything above is evidence. This section is what one resident can actually do with it between now and the contract decision. Two things matter more than volume: getting the question in front of the full Council in a form that produces a recorded vote, and being accurate enough that nothing you send can be corrected in public.

Who to write, with verified addresses

Every address below was taken from the Town of Apex website and staff directory on September 3, 2026. Titles are the Town’s own.

• Anything you send to a government address is a public record under Chapter 132 of the North Carolina General Statutes. That is not a reason to hold back. It is a reason to write something you would be content to see quoted, and it is also why a written record of resident contact matters more than a phone call.
• The Public Safety Committee is two people: Mayor Gilbert as Chair and Council Member Ed Gray. Committee meetings are quarterly and open to the public, with notice on the Town calendar.
• Check the committee date before you plan around it. The meeting on the plate reader recommendation was originally scheduled for September 24, 2026 and is reported to have been moved by Mayor Gilbert to October 1, 2026. That change was not confirmed as of this writing. Confirm it yourself against the Town calendar at apexnc.org/Calendar.aspx and the agendas page at apexnc.org/838/Agendas-Minutes once the notice is posted, or call the Town Clerk at (919) 249-1260. Do not repeat either date in public as settled until you have seen the posted notice.
• Ask one question the Town has not answered: was the addendum executed? The Town’s own released email shows staff routing a Law Enforcement Agreement in April 2026 to add ten more cameras to the existing ten, against a purchase order deadline. The records do not show whether it was signed. If it was, the Council is being asked to renew a program twice the size of the one it approved in January 2025. Ask for the signed document and the purchase order. See Section 6.
• Regular Town Council meetings are the 2nd and 4th Tuesday of each month at Town Hall, 73 Hunter Street. Work sessions are the 3rd Tuesday. December sometimes runs a single meeting, so confirm the date with the Clerk before you plan around it.

Why Mayor Gilbert is the most important person to write, even though he does not cast a vote

Mayor Gilbert will not be one of the votes counted on this. He is nonetheless the person with the most control over whether a vote happens at all, and the person whose position will carry the most weight with everyone else in the room.

• He chairs the Public Safety Committee, the two member committee where the plate reader recommendation is formed before it reaches the full Council.
• He sets the committee assignments. The Town’s own page states that committee and liaison assignments are set by the Mayor every two years in collaboration with the Council. The current assignments were announced at the April 21, 2026 work session.
• His background is why his voice outweighs a vote. He served in the Apex Police Department from 1990 to 2019 and retired as a Police Captain, twenty nine years. He is a graduate of the FBI National Academy, 220th Session. Since 2022 he has been a Commissioner on the North Carolina Justice Education and Training Standards Commission, the state body that sets law enforcement training and certification standards. He chairs the North Carolina Metropolitan Mayors Coalition, founded Blue Lights College, and graduated from Apex High School.
• Read that list again from the Department’s point of view. There is no one in Apex better positioned to tell the Police Department that a recorded public vote is good for them. If the mayor who spent twenty nine years in that department asks the Council to defend this program in public, and the Council does, the program has a legitimacy that a consent agenda cannot give it.

The ask: one clean, vendor-neutral, recorded vote

The specific request, in one sentence you can paste into any email:

• Why it has to be standalone. If the motion that reaches the Council is “drop Flock, add Axon,” then the Council votes on a vendor and never votes on the technology. The four options in Section 13 only exist if the question is asked properly.
• Why it has to be recorded. The program was authorized on January 14, 2025 on the consent agenda, where a council puts items it does not expect to debate. The vote was 5 to 0 and there was no stand-alone discussion and no public hearing. Nobody in Apex has ever had to explain a position on this in public.
• Why vendor-neutral. Every governance term in Section 13 applies to any vendor. And Section 7 documents the mechanism that makes a vote skippable: plate readers are bought as a service subscription, which North Carolina’s competitive bidding statutes do not reach at any dollar figure, so whether the Council votes is set by the Town’s own purchasing policy rather than by state law. A vendor added by amendment to an existing master agreement, or a $0 pilot, produces no vote at all.
• Why it is not a request to vote no. It is a request that the decision be made the way decisions of this size are supposed to be made. A council member who supports the cameras loses nothing by voting for them on the record. A resident who opposes them at least knows who to hold responsible. Everyone is better off than they are under a consent agenda.

How to write something that actually lands

• Use your own words in the subject line and the first paragraph. Everything else can be pasted. This single step is the difference between a counted constituent contact and a form letter.
• Keep it under about 300 words and make one ask per email.
• Write from your own home address and say where in Apex you live. Constituency is the whole basis of the request.
• Do not attack the Police Department. The strongest form of this argument is that the Department is being asked to carry a governance and liability risk that the Town has not managed, and that is a Council problem rather than a police problem.
• Stay inside the verified record. Appendix B lists five claims circulating locally that do not survive checking, including a Fox News poll that does not appear to exist and the claim that Trump owns Axon. Using any of them will cost you the room.
• Ask for a reply, and keep track of who answers. Who responded, and what they said, is itself useful information for the next email and for the press.
• Then show up. A room with new faces in it is worth more than the twentieth email from the same ten people. Ask the Clerk how to sign up to speak, and bring one specific point rather than the whole briefing.

A two week cadence from November 1

Three emails, spaced two weeks apart, each landing a few days before a scheduled Council meeting. Copy each one, change the top, send it.

If you want to do more than email

1. Speak at a regular meeting. 2nd and 4th Tuesday, 73 Hunter Street. Contact the Town Clerk to confirm the agenda and the sign-up procedure. One specific point, ninety seconds, your own experience.
2. Attend a Public Safety Committee meeting. It is a two person committee that meets quarterly, open to the public, noticed on the Town calendar. This is where the recommendation is written. Confirm the date first: the plate reader meeting was set for September 24, 2026 and is reported to have moved to October 1, and neither has been confirmed. Check apexnc.org/Calendar.aspx once the notice is posted.
3. File a public records request. The Town’s Flock audit log, the current network sharing roster, the Department’s Article 3D written policy, and the annual audit required by G.S. 20-183.31(a)(7) are all public records. Appendix C lists what is still unknown and which document settles each item.
4. Ask your neighbors to write in their own words. Six personalized emails from six different streets outweigh sixty identical ones.
5. Send this briefing to one person who disagrees with you. Section 3 makes the case for the cameras at full strength and names the rescues. If the argument only works on people who already agree, it is not an argument.

App AQuestions for the vendor, the Chief and the Town Attorney

Each question below is answerable, specific, and traceable to a documented failure elsewhere in this briefing. None of them require anyone to concede a position.

• If the recommendation is Axon, ask these four before any vote. One: will the Town be enrolled in the AI Era Plan or any AI report writing product, and at what per-officer cost, and can that be excluded in writing? Two: has the Wake County District Attorney confirmed in writing that this office will accept police reports drafted with AI assistance, given that the King County, Washington prosecutor refuses them? Three: which of the fifteen recommendations in Axon’s own former AI Ethics Board plate reader report does this contract satisfy? Four: what is retained, and for how long, when an AI-generated report describes an incident in which no crime occurred and a minor is involved? See Sections 11.8 and 11.9.

For the Police Department

1. Publish the Austin chain for Apex, for the full contract term: plates scanned, alerts generated, stops made, stops released because of a data error, arrests, prosecutions, convictions.
2. How many of the arrests attributable to the system involved a felony against a person, and how many involved shoplifting, fraud or property offenses along NC 55?
3. What is the Department’s alert level error rate, not the vendor’s per scan accuracy claim? If the Department does not measure it, when will it start?
4. Is independent visual verification of the plate mandatory in writing before any stop based on an alert? If so, produce the policy. If not, why not, given the Green, Gilliam, Morristown and Sherwood cases?
5. Is a suspected stolen vehicle a high risk stop under Apex policy, with weapons drawn as a matter of course?
6. Reconcile the three federal searches the Town reported with the 9,683 reported by DeFlock Apex, with a written definition of each term.
7. Produce the complete Flock audit log for January 1 through August 31, 2026, and the current network sharing roster: every organization with access, in state and out.
8. Who at the Town enabled the Postal Inspection Service and the Air Force base police department, when, under what authority, and what control now prevents a repeat?
9. How many searches has the Department run, how many had a documented case number, and who has reviewed a sample of them against the underlying case files?
10. Produce the Department’s Article 3D written policy and the annual audit and report of system use and effectiveness required by G.S. 20-183.31(a)(7).

For the vendor, whichever vendor

1. Reconcile your published accuracy specification with your marketing accuracy claim, and state the alert level error rate for deployments configured like Apex’s.
2. Which capabilities can be added to this contract without returning to this Council for a vote? Name each one. Audio detection, drones, plate-free vehicle matching, people lookup and private hotlists have all been added elsewhere without a vote.
3. Provide the redlined master agreement, not the FAQ or the blog. Specifically: the limitation of liability, the indemnification clause, the warranty disclaimer, and any license you retain to use, disclose or train models on Town data.
4. Will you accept an accuracy standard with a remedy, such as withheld payment or termination on failure to meet a measured alert level threshold?
5. Will you accept the Dunwoody terms: a data security indemnity, a bar on using Town data to train machine learning models without written authorization, and $5 million each in cyber and professional liability coverage?
6. Will you accept independent third party verification of deletion at your storage environment, funded by the Town, by an auditor with no financial relationship to you?
7. Will you accept termination for convenience on 30 days notice with no penalty and certified data deletion, given that the Fourth Circuit is likely to rule on this technology within the contract term?
8. Confirm in writing every federal contract, pilot and data sharing arrangement you currently hold, and every one you held in the past 24 months.
9. For Axon specifically: confirm in writing that the Town is opted out of the Axon Customer Experience Improvement Program, and state which of your own AI Ethics Board’s fifteen ALPR recommendations this contract satisfies.
10. For Axon specifically: is Fusus included, at what price, and what is the complete list of camera feeds it would be able to access on day one?

For the Town Attorney and the insurance pool

1. Does Apex carry the optional law enforcement liability and cyber and privacy coverage lines through the League pool, at what limits, and with what retention?
2. Does the general liability form contain an invasion of privacy or statutory violation exclusion that would defeat coverage for a privacy claim of the kind now pleaded against Flock in California?
3. Will the pool confirm in writing that ALPR deployment does not change coverage terms, and price the change if it does?
4. What is the Town’s exposure under 42 U.S.C. 1983 and 1988 for a wrongful stop, given that G.S. 160A-485’s policy limit cap does not apply to a federal claim?
5. Can the Town impose retention shorter than the state’s 90 day ceiling on its own department under G.S. 160A-174(b)’s higher standards clause? Can it regulate private ALPR siting? Both questions are legally untested in North Carolina and need an opinion, not an assumption.
6. What is the total cost of the 2024 cyber incident, including forensics, the remediation firm, the replacement billing system, legal fees, notification, credit monitoring, staff overtime and written off revenue?
7. Does the Town’s purchasing policy currently require a Council vote on a service subscription of this size, and on a $0 pilot? If not, should it?

App BFive claims to stop using

Each of these is circulating in the current local debate. Each is wrong or unsupported. Each has a stronger, verifiable replacement. The arguments in this briefing do not need any of them, and using one hands the other side an easy correction in front of the room.

1. The Fox News poll showing single digit and teens support

• The problem: Fox News’s own official poll archive contains no poll on license plate readers, ALPR or government surveillance. The cited figures, Republicans in the teens and Democrats around 11 percent, are inconsistent with every poll that does exist, and no poll shows Democrats less supportive than Republicans on this issue.
• Say instead: Knox County, Tennessee, August 2026, Spry Strategies, 350 likely voters, 58.3 percent opposed to police use of Flock cameras. Arizona statewide, August 2026, NextGen Polling, 1,627 registered voters, 63 percent opposed and 82 percent concerned about data misuse. Both are real, recent, sourced and sufficient.

2. Trump and other right wing figures are Axon investors

• The problem: a $1 million to $5 million position against a $48.9 billion market capitalization is about 0.002 to 0.010 percent of the company, hundreds of times below the threshold that would require an ownership filing. There is no Trump Schedule 13D or 13G on Axon. Donald Trump Jr. is not on Axon’s board or advisory board, though he holds roles at ten other companies. Axon’s actual owners are index funds, and its political giving is close to evenly split between the parties.
• Say instead: a federal ethics filing shows the President personally purchased between $1 million and $5 million of Axon stock on February 10, 2026, and ICE posted a roughly $220 million solicitation matching Axon’s product two weeks later. That is a federal conflict of interest question, and it is a different claim from ownership. Add that both vendors are backed by mainstream growth capital, and that the more politically prominent investor is actually on Flock’s side of the ledger.

3. Denver signed a $498,500 extension to come in $1,500 under the vote threshold

• The problem: this figure appears on an advocacy contract tracker, could not be corroborated, and conflicts with contemporaneous local reporting describing a no cost extension.
• Say instead: Denver’s council unanimously rejected a $666,000 renewal in May 2025. In October 2025 the mayor extended the contract at no cost through March 2026, and because there was no cost, council approval was not required. A council majority publicly called it undemocratic. A $0 pilot is the most effective vote avoidance structure documented anywhere in this research, because it defeats both the dollar threshold and the cost objection at once. The verified version is the stronger argument.

4. Flock faces an imminent North Carolina licensing ban

• The problem: the documented case is from October 2023, it concerned alarm systems installer licensing under Chapter 74D rather than private investigator licensing, and Flock obtained a license in July 2024 and now serves hundreds of North Carolina customers. Whether any live 2026 licensing action threatens Flock’s operations here could not be verified.
• Say instead: in October 2023 a Wake County Superior Court judge ordered Flock to stop installing systems in North Carolina unless it used a licensed third party, after the state’s Alarm Systems Licensing Board found it was operating unlicensed, and the court found the unlicensed work “an immediate and direct threat to the public health, safety and welfare.” At least eight North Carolina agencies were notified their contracts were likely void. That is a documented history of the company operating outside a state licensing requirement until a court stopped it, which is a fair point about vendor conduct and does not require claiming a current ban.

5. Flock will never delete the data, and once it is on the internet it never leaves

• The problem: the strong version is not established, and a vendor representative will rebut it in one sentence by pointing at the retention setting.
• Say instead: deletion is a vendor side event that the Town cannot observe or verify. Retention is a configurable setting, and an LAPD audit surfaced a Flock contract permitting five years. Search results, exports and case files leave the platform and follow other agencies’ retention rules, and roughly 7,000 agencies can reach a participating department’s data, with no mechanism by which Apex can recall a record already shared. In Littleton, Massachusetts, the town voted in May 2026 to shut its cameras down and Flock restored power to five of six in August, saying it had not been told the shutdown was permanent. The defensible sentence is that plate data, once collected and shared, is outside the Town’s control, cannot be recalled, and the Town has no way to confirm it was destroyed.

App COpen questions and the document that settles each

Sixteen items could not be verified for this briefing. Each is answerable, and most are answerable in a day. The six drawn from the Town’s own released email are marked in Section 6.

App DSourcing note

• Appendix E is the full source list, with a link to everything cited in this briefing, grouped into primary law and case law, North Carolina statutes and procurement rules, government audits and enforcement actions, peer reviewed and academic research, vendor materials and executed contracts, journalism, United Kingdom and European sources, Apex and Wake County sources, and the advocacy compiled trackers. Appendix E13 lists all 194 records captured from the Institute for Justice abuse database, record by record, with the source the database cites for each one.
• The Town’s own records. Section 6 rests on Town of Apex email released under the North Carolina Public Records Act and on a spreadsheet prepared by the Police Department for its Chief. Quotations are transcribed from the released documents. Several threads are redacted at the point of the operative answer, and where that is so, Section 6 says what the record does not show rather than inferring it. How this briefing was built. Every factual claim was traced to a primary document, a government report, a peer reviewed study, a named news outlet, or a labeled vendor or advocacy source. Sources were read rather than cited from memory. The abuse database in Section 5 was scraped and tabulated directly rather than taken from a summary, which is why the figures in this briefing differ from the round numbers that circulate in press coverage of it.
• Currency of this briefing. First prepared September 2, 2026; this is the third major revision, dated September 4, 2026, and it is updated as new records and data are collected. Where a figure depends on a date, the date is given, because several of them will move: the abuse database grows, records requests are still outstanding, and the vendor changed its own terms twice in the twelve months before this was written. Method and limits. Where two sources conflicted, both figures appear with the conflict named. Where a claim could not be traced to a primary or major source, it was either omitted or explicitly flagged. Where a widely repeated advocacy claim did not survive checking, it appears in Appendix B rather than in the body. Statistics attributed to Flock or Axon are the companies’ own and are labeled that way, because in most cases no independent verification is possible. Counts drawn from the abuse database are stated as “at least,” because the database itself is an incomplete record of an under-detected problem. Nothing in this briefing should be quoted publicly without the two verifications listed at the top of Appendix C being completed first.

App EFull source list

Every source relied on in this briefing, grouped by type. URLs are given without the protocol prefix to keep them readable in print. Where a source is a vendor or an advocacy organization, that is stated in the label, because the label is what a reader will see if this list is quoted.

E1. The abuse database and the litigation behind it

Institute for Justice, The IJ Database of ALPR Abuse (advocacy; tabulated for Section 5 on September 3, 2026). ij.org/the-ij-database-of-alpr-abuse/

Institute for Justice, Dozens of Innocent Motorists Have Been Pulled Over, Detained at Gunpoint, or Jailed Due to AI License Plate Camera Errors (advocacy). ij.org/dozens-of-innocent-motorists-have-been-pulled-over-detained-at-gunpoint-or-jailed-due-to-ai-license-plate-camera-errors/

Institute for Justice, stalking incident tracker (advocacy). ij.org/police-have-reportedly-used-license-plate-readers-to-stalk-romantic-interests-at-least-14-times-in-recent-years/

Institute for Justice, press release on the Norfolk suit proceeding (advocacy). ij.org/press-release/judge-rules-lawsuit-challenging-norfolks-use-of-flock-cameras-can-proceed/

Institute for Justice, court rejects Flock’s bid to intervene in Norfolk (advocacy). ij.org/press-release/federal-court-rejects-flock-safetys-late-bid-to-join-and-block-ijs-lawsuit-challenging-norfolks-mass-surveillance-cameras/

Kansas Justice Institute, ALPR case page, Wichita (advocacy). kansasjusticeinstitute.org/case/automated-license-plate-readers/

Electronic Frontier Foundation, The Human Toll of ALPR Errors (advocacy). eff.org/deeplinks/2024/11/human-toll-alpr-errors

E2. Constitutional and case law

Katz v. United States and the third-party doctrine, Congressional Research Service R43586. everycrsreport.com/reports/R43586.html

United States v. Knotts, 460 U.S. 276 (1983). supreme.justia.com/cases/federal/us/460/276/

New York v. Class, 475 U.S. 106 (1986). caselaw.findlaw.com/court/us-supreme-court/475/106.html

United States v. Jones, 565 U.S. 400 (2012). supreme.justia.com/cases/federal/us/565/400/

Carpenter v. United States, 585 U.S. 296 (2018), Cornell LII. law.cornell.edu/supremecourt/text/16-402

Chatrie v. United States, slip opinion, June 29, 2026. supremecourt.gov/opinions/25pdf/25-112_0am4.pdf

Chatrie, SCOTUSblog case report. scotusblog.com/2026/06/court-rules-that-law-enforcements-use-of-geofence-warrant-was-a-search/

Chatrie analysis, WilmerHale client alert. wilmerhale.com/en/insights/client-alerts/20260702-the-fourth-amendment-moves-to-the-cloud

Chatrie, ACLU statement (advocacy). aclu.org/press-releases/aclu-applauds-important-supreme-court-decision-making-clear-location-data-is-protected-by-the-constitution

Leaders of a Beautiful Struggle v. Baltimore Police Dept., 4th Cir. en banc, 2021. law.justia.com/cases/federal/appellate-courts/ca4/20-1495/20-1495-2021-06-24.html

Commonwealth v. McCarthy, Mass. SJC, 2020. law.justia.com/cases/massachusetts/supreme-court/2020/sjc-12750.html

Schmidt v. City of Norfolk ruling, Courthouse News. courthousenews.com/judge-holds-norfolks-license-plate-reader-use-constitutional/

Schmidt v. City of Norfolk ruling, WHRO. whro.org/business-growth/2026-02-11/a-federal-judge-ruled-norfolks-flock-surveillance-cameras-dont-invade-peoples-privacy-yet

Schmidt appeal status, Fourth Circuit No. 26-1227. recordinglaw.com/news/norfolk-flock-license-plate-cameras-fourth-amendment-appeal/

EPIC amicus brief, Schmidt v. Norfolk. epic.org/wp-content/uploads/2026/04/EPIC-Amicus-Schmidt-v.-Norfolk.pdf

Commonwealth v. Bell and Church, EFF analysis (advocacy). eff.org/deeplinks/2025/09/eff-urges-virgina-court-appeals-require-search-warrants-access-alpr-databases

State v. Simonson, Washington courts filing. courts.wa.gov/content/petitions/1052375%20Answer%20to%20Petition%20for%20Review.pdf

Idris v. City of Chicago, 552 F.3d 564 (7th Cir. 2009). openjurist.org/552/f3d/564

42 U.S.C. 1983, Cornell LII. law.cornell.edu/uscode/text/42/1983

42 U.S.C. 1988, Cornell LII. law.cornell.edu/uscode/text/42/1988

Chatrie’s implications for plate readers, Ohio Capital Journal. ohiocapitaljournal.com/2026/07/08/are-flock-cameras-next-supreme-court-rules-against-police-using-sprawling-cell-phone-location-data/

E3. North Carolina law, procurement and the state highway program

N.C. Gen. Stat. Chapter 20, Article 3D, sections 20-183.30 to 20-183.33. ncleg.gov/EnactedLegislation/Statutes/HTML/ByArticle/Chapter_20/Article_3D.html

G.S. 20-183.32, the 90 day retention cap. ncleg.net/enactedlegislation/statutes/html/bysection/chapter_20/gs_20-183.32.html

UNC School of Government, license plate readers, the highway pilot program and the new offense. nccriminallaw.sog.unc.edu/license-plate-readers-highway-pilot-program-and-new-offense/

UNC School of Government, what you might have missed in the 2026 budget. nccriminallaw.sog.unc.edu/2026/07/27/things-you-might-have-missed-in-the-budget/

Senate Bill 257, Session Law 2026-41, the 2026 Appropriations Act. ncleg.gov/BillLookup/2025/S257

Carolina Journal, budget makes plate readers permanent. carolinajournal.com/nc-budget-makes-license-plate-readers-permanent-amid-privacy-concerns/

WRAL, state highway plate reader program made permanent. wral.com/news/nccapitol/north-carolina-highway-license-plate-reader-permanent-june-2026/

WECT investigates, state plate reader program expands. wect.com/2026/07/21/wect-investigates-state-license-plate-reader-program-expands/

G.S. 160A-174, limits on municipal ordinances and the higher standards clause. law.justia.com/codes/north-carolina/chapter-160a/article-8/section-160a-174/

G.S. 74D-2, Security Systems Licensing Act. ncleg.gov/EnactedLegislation/Statutes/HTML/BySection/Chapter_74D/GS_74D-2.html

GovTech, North Carolina judge halts ALPR deployment over licensing. govtech.com/public-safety/north-carolina-judge-halts-alpr-deployment-over-licensing

G.S. 160A-485, waiver of governmental immunity by purchase of insurance. ncleg.gov/EnactedLegislation/Statutes/PDF/BySection/Chapter_160a/GS_160a-485.pdf

G.S. 160A-300.1, traffic control photographic systems. ncleg.gov/EnactedLegislation/Statutes/HTML/BySection/Chapter_160A/GS_160A-300.1.html

G.S. 115C-437, clear proceeds of fines and forfeitures. ncleg.gov/EnactedLegislation/Statutes/HTML/BySection/Chapter_115C/GS_115C-437.html

G.S. 143-129, formal bidding. ncleg.gov/EnactedLegislation/Statutes/PDF/BySection/Chapter_143/GS_143-129.pdf

G.S. 143-131, informal bidding. ncleg.net/EnactedLegislation/Statutes/HTML/BySection/Chapter_143/GS_143-131.html

G.S. 143-133, no evasion permitted. law.justia.com/codes/north-carolina/chapter-143/article-8/section-143-133/

North Carolina contracting dollar thresholds table. abc.nc.gov/media/1303/open

A representative North Carolina local purchasing policy, showing service contracts exempt from bidding. alexandercountync.gov/pdf/purchasing_policy.pdf

G.S. 143-318.12, open meetings notice. law.justia.com/codes/north-carolina/chapter-143/article-33c/section-143-318-12/

G.S. 143B-1379, local government cyber incident reporting. ncleg.gov/EnactedLegislation/Statutes/HTML/BySection/Chapter_143B/GS_143B-1379.html

UNC School of Government, key North Carolina cybersecurity legislation, including the G.S. 143-800 ransom payment ban. sog.unc.edu/sites/default/files/cyber_legislation.pdf

UNC School of Government Coates’ Canons, Fearrington v. City of Greenville. canons.sog.unc.edu/blog/2024/07/11/fearrington-v-city-of-greenville-north-carolina-supreme-court-reverses-court-of-appeals-and-upholds-citys-red-light-camera-enforcement-program-as-constitutional/

UNC School of Government, Greenville’s red light camera program ruled unconstitutional. nccriminallaw.sog.unc.edu/2022/03/28/greenvilles-red-light-camera-program-ruled-unconstitutional/

NCDOT encroachment agreements. connect.ncdot.gov/municipalities/Utilities/pages/encroachment-agreements.aspx

North Carolina State Bureau of Investigation crime statistics. ncsbi.gov/Services/Crime-Statistics

E4. Government audits, official reports and enforcement actions

California State Auditor, Report 2019-118, Automated License Plate Readers, February 2020. information.auditor.ca.gov/pdfs/reports/2019-118.pdf

Los Angeles Police Department Inspector General ALPR audit, July 2026, reported by NBC Los Angeles. nbclosangeles.com/investigations/lapd-watchdog-urges-reconsideration-of-automatic-license-plate-readers/3916867/

Austin City Auditor, ALPR program audit, 2024. services.austintexas.gov/edims/document.cfm?id=451701

Austin audit summary, Community Impact. communityimpact.com/austin/south-central-austin/government/2025/05/20/75m-license-plates-scanned-under-rebooted-austin-police-program-audit-reveals-successes-concerns/

Seattle Office of Inspector General, 2024 ALPR surveillance audit. seattle.gov/documents/departments/oig/audits/december%202024/oig_2024surveillance%20report_alpr.pdf

Virginia State Crime Commission, 2024 annual report on law enforcement use of ALPR. vscc.virginia.gov/Annual%20Reports/2024%20VSCC%20Annual%20Report%20-Law%20Enforcement%20Use%20of%20ALPR.pdf

Virginia State Crime Commission, HB 2724 fact sheet. vscc.virginia.gov/2024/VSCC%20HB%202724%20ALPR%20Fact%20Sheet.pdf

Virginia Code 2.2-5517, ALPR access, logging and retention. law.lis.virginia.gov/vacode/title2.2/chapter55.1/section2.2-5517/

Illinois Secretary of State audit findings, Capitol News Illinois. capitolnewsillinois.com/news/illinois-license-plate-cameras-used-illegally-by-out-of-state-police-giannoulias-says/

Illinois audit and Flock’s response, WTTW Chicago. news.wttw.com/2025/08/26/license-plate-camera-company-operating-illinois-halts-cooperation-federal-agencies-amid

Forest Park, Illinois audit finding, Chicago Sun-Times. chicago.suntimes.com/immigration/2025/09/05/state-audit-forest-park-license-plate-data-ice-alexi-giannoulias

University of Washington Center for Human Rights, Leaving the Door Wide Open, October 2025. jsis.washington.edu/humanrights/2025/10/21/leaving-the-door-wide-open/

California Attorney General sues El Cajon over out of state ALPR sharing. oag.ca.gov/news/press-releases/attorney-general-bonta-sues-el-cajon-illegally-sharing-license-plate-data-out

DHS Office of Inspector General, OIG-20-71, review of the CBP Perceptics incident. oig.dhs.gov/reports/2020/review-cbps-major-cybersecurity-incident-during-2019-biometric-pilot/oig-20-71-sep20

FBI Uniform Crime Reporting, Summary of Reported Crimes in the Nation, 2024. cde.ucr.cjis.gov/LATEST/resources/reports/UCR%20Summary%20of%20Reported%20Crimes%20in%20the%20Nation%202024.pdf

Policing Project at NYU School of Law, ALPR fifty state scorecard, August 2026. policingproject.org/automated-license-plate-readers

Fifty state scorecard coverage, GovTech. govtech.com/biz/data/alpr-policies-all-50-states-ranked-on-privacy-data-sharing

Senator Josh Hawley, investigation of Flock Safety, August 26, 2026. hawley.senate.gov/chairman-hawley-investigates-ai-powered-flock-cameras-amidst-privacy-concerns/

United States Senate Lobbying Disclosure Act filings, Axon Enterprise, 2025. lda.gov/api/v1/filings/?client_name=Axon+Enterprise&filing_year=2025

Office of Government Ethics Form 278-T, Donald J. Trump, filed May 2026, disclosing the Axon purchase. static.notus.org/2b/18/a1761d44459fb1c67e196c1b3629/trump-donald-j-05-08-2026-278t2.pdf

E5. Peer reviewed and academic research

Lum, Hibdon, Cave, Koper and Merola (2011), License plate reader police patrols in crime hot spots, Journal of Experimental Criminology. link.springer.com/article/10.1007/s11292-011-9133-9

George Mason University Center for Evidence-Based Crime Policy, license plate reader evidence review. cebcp.org/wp-content/evidence-based-policing/LPR_FINAL.pdf

Koper and Lum (2019), The Impacts of Large-Scale License Plate Reader Deployment on Criminal Investigations, Police Quarterly, via the National Institute of Justice. nij.ojp.gov/library/publications/impacts-large-scale-license-plate-reader-deployment-criminal-investigations

Adams and Mourtgos (2026), Automated License Plate Readers, Vehicle Theft, and Clearance, CrimRxiv working paper, not peer reviewed. crimrxiv.com/pub/zleg04q3/release/1

Colorado Division of Criminal Justice, ALPR research brief. dcj.colorado.gov/dcj-offices/ors/doc-res/brf-alpr2

Pew Research Center, what the data says about crime in the United States. pewresearch.org/short-reads/2024/04/24/what-the-data-says-about-crime-in-the-us/

Jeff Asher, analysis of FBI clearance data. jasher.substack.com/p/lagniappe-thoughts-on-the-fbis-annual

Atlanta Community Press Collective, Flock cameras and Atlanta clearance rates, July 2026. atlpresscollective.com/2026/07/28/flock-cameras-atlanta-clearance-rates/

E6. Vendor materials and contracts

Flock Safety, ALPRs and the Fourth Amendment, the company’s own legal position (vendor). flocksafety.com/blog/automated-license-plate-readers-and-the-fourth-amendment-a-public-safety-by-design-perspective-from-flock

Flock Safety, How Effective Is Flock, 2025 Impact Census (vendor). flocksafety.com/blog/how-effective-is-flock

Flock Safety, How many crimes do ALPRs solve anyway (vendor). flocksafety.com/customers/how-many-crimes-do-automated-license-plate-readers-alprs-solve-anyway

Flock Safety, 2024 press release claiming 10 percent of reported crime (vendor). globenewswire.com/news-release/2024/02/08/2826075/0/en/New-Study-Finds-that-Flock-Safety-is-Instrumental-in-Solving-10-of-Reported-Crime-in-U-S.html

Flock Safety, Does Flock share data with ICE (vendor). flocksafety.com/blog/does-flock-share-data-with-ice

Flock Safety, statement on network sharing and federal cooperation (vendor). flocksafety.com/blog/statement-network-sharing-use-cases-federal-cooperation

Flock Safety, who really has access to your Flock data (vendor). flocksafety.com/blog/who-really-has-access-to-your-flock-data

Flock Safety, August 13, 2026 privacy, accountability, security and transparency safeguards (vendor). flocksafety.com/blog/flock-guardrails-address-lpr-privacy-concerns-and-police-transparency

Flock Safety, August 21, 2026 updates to terms and conditions (vendor). flocksafety.com/blog/building-on-trust-updates-to-flocks-terms-and-conditions

Flock Safety, response to compiled security research on Flock devices (vendor). flocksafety.com/blog/response-to-compiled-security-research-on-flock-safety-devices

Flock Safety, Has Flock been hacked (vendor). flocksafety.com/blog/has-flock-been-hacked

Flock Safety, HOA marketing page (vendor). flocksafety.com/industries/hoas

Flock Safety, promotion of the 11 percent vehicle theft study (vendor). flocksafety.com/blog/independent-study-vehicle-thefts-fell-11-after-flock-cameras-went-live

Flock Safety Master Services Agreement executed by Groton, Massachusetts, standard form. portal.grotonma.gov/storage/Police/ALPR/Flock_Agreement_2.pdf

Flock Safety Master Services Agreement executed by Dunwoody, Georgia, heavily negotiated. haveibeenflocked.com/blog/dunwoody-2026-04-13/Flock_Master_Services_Agreement.pdf

Axon Master Services and Purchasing Agreement, version 1.0 (vendor). a.storyblok.com/f/198504/x/e9c5ffd778/mspa_v1-0.pdf

Axon Outpost fixed plate reader product page (vendor). axon.com/products/axon-outpost

Axon Lightpost streetlight retrofit product page (vendor). axon.com/products/axon-lightpost

Axon Fusus real time crime center product page (vendor). axon.com/products/axon-fusus

Axon Draft One product page (vendor). axon.com/products/draft-one

Axon privacy policy (vendor). axon.com/privacy

Axon Trust Center, CJIS and FedRAMP posture (vendor). trust.axon.com/

Axon second quarter 2026 results, including recurring revenue and contracted backlog. investor.axon.com/2026-08-05-Axon-reports-Q2-2026-revenue-of-904-million,-up-35-year-over-year

Axon acquisition of Fusus, February 2024. investor.axon.com/2024-02-01-Axon-Accelerates-Real-Time-Operations-Solution-with-Strategic-Acquisition-of-Fusus

Axon acquisition of Dedrone, May 2024. investor.axon.com/2024-05-06-Axon-to-acquire-Dedrone,-accelerating-the-next-generation-of-drone-solutions-to-protect-more-lives-in-more-places

Axon business combinations note, SEC filing. sec.gov/Archives/edgar/data/1069183/000106918325000077/R23.htm

Axon 2026 proxy statement, DEF 14A. sec.gov/Archives/edgar/data/1069183/000162828026025521/axon-20260416.htm

Policing Project, the Axon AI Ethics Board. policingproject.org/axon

Axon AI Ethics Board first report, on face recognition. policingproject.org/axon-fr

Axon AI Ethics Board report on ALPR, with fifteen recommendations. policingproject.org/axon-alpr

E7. Journalism and investigations

Forbes, Flock AI cameras may not reduce crime, February 2024. forbes.com/sites/cyrusfarivar/2024/02/29/flock-ai-cameras-may-not-reduce-crime/

Forbes, Axon versus Flock, the falling out, February 2025. forbes.com/sites/thomasbrewster/2025/02/07/axon-vs-flock-two-policing-tech-giants-just-had-a-big-falling-out/

404 Media, researcher who oversaw the Flock study now has concerns about it. 404media.co/researcher-who-oversaw-flock-surveillance-study-now-has-concerns-about-it/

404 Media, ICE taps into a nationwide AI enabled camera network. 404media.co/ice-taps-into-nationwide-ai-enabled-camera-network-data-shows/

404 Media, Flock removes states from the national lookup tool. 404media.co/flock-removes-states-from-national-lookup-tool-after-ice-and-abortion-searches-revealed/

404 Media, a Texas officer searched cameras nationwide for a woman who got an abortion. 404media.co/a-texas-cop-searched-license-plate-cameras-nationwide-for-a-woman-who-got-an-abortion/

404 Media, CBP had access to more than 80,000 Flock cameras. 404media.co/cbp-had-access-to-more-than-80-000-flock-ai-cameras-nationwide/

EFF, Flock and a Texas sheriff claimed a search was for a missing person, the records say otherwise (advocacy). eff.org/deeplinks/2025/10/flock-safety-and-texas-sheriff-claimed-license-plate-search-was-missing-person-it

EFF, policy position on ALPR surveillance, August 2026 (advocacy). eff.org/deeplinks/2026/08/effs-policy-position-alpr-surveillance-eliminate-it-and-reduce-its-harms

EFF, Axon’s Draft One is designed to defy transparency (advocacy). eff.org/deeplinks/2025/07/axons-draft-one-designed-defy-transparency

EFF, new ALPR vulnerabilities in Motorola equipment (advocacy). eff.org/deeplinks/2024/06/new-alpr-vulnerabilities-prove-mass-surveillance-public-safety-threat

EFF, license plate readers exposed on the open internet, 2015 (advocacy). eff.org/deeplinks/2015/10/license-plate-readers-exposed-how-public-safety-agencies-responded-massive

EFF, community control over police surveillance ordinances (advocacy). eff.org/deeplinks/2021/05/community-control-police-spy-tech

ACLU, Flock Safety’s credibility lost as it repeatedly misstates facts to city councils (advocacy). aclu.org/news/privacy-technology/tracking-alpr-cameras/flock-safety-credibility-lost-as-it-repeatedly-lies-to-city-councils-police-departments-and-public-across-the-country

ACLU, Flock’s aggressive expansions go far beyond driver surveillance (advocacy). aclu.org/news/privacy-technology/tracking-alpr-cameras/flock-roundup

ACLU, despite new updates Flock’s cameras remain a civil liberties threat, August 2026 (advocacy). aclu.org/news/privacy-technology/tracking-alpr-cameras/despite-new-updates-flocks-creepy-cameras-remain-major-civil-liberties-threat

ACLU of Massachusetts, Flock gives law enforcement nationwide access to your location (advocacy). data.aclum.org/2025/10/07/flock-gives-law-enforcement-all-over-the-country-access-to-your-location/

ACLU, the perpetual license in Flock’s standard agreement (advocacy). aclu.org/news/privacy-technology/flock-massachusetts-and-updates

ACLU, Denise Green, pulled from her car at gunpoint over a misread plate (advocacy). aclu.org/news/privacy-technology/san-francisco-woman-pulled-out-car-gunpoint-because

Techdirt, studies show Flock’s ALPRs reduce crime, so long as Flock controls the inputs. techdirt.com/2024/04/01/studies-show-flocks-alprs-reduce-crime-so-long-as-flock-controls-the-inputs-and-the-methodology/

Techdirt, Flock reactivates a camera system a town had voted to shut down. techdirt.com/2026/08/13/flock-again-activates-a-camera-system-a-town-had-voted-to-shut-down/

Boston Globe, Littleton discovers Flock restored power to disabled cameras. bostonglobe.com/2026/08/06/business/flock-littleton-license-plate-camera-disable/

Fortune, Flock’s rise, funding, cancellations and the five year retention contract. fortune.com/2026/08/24/flock-8-billion-startup-backed-a16z-tiger-surveillance-camera/

Stateline, Flock tightens safeguards as states and cities question the network. stateline.org/2026/08/14/flock-safety-tightens-safeguards-as-states-cities-question-surveillance-network/

NBC News, Virginia lawsuit over a driver tracked 526 times. nbcnews.com/tech/security/virginia-police-used-flock-cameras-track-driver-safety-lawsuit-surveil-rcna230399

Gizmodo, Flock cameras got the wrong plate 71 percent of the time in one California city. gizmodo.com/flock-cameras-got-the-wrong-license-plate-71-of-the-time-in-california-city-2000793483

Gizmodo, the Florida memo and its impact on Flock cameras. gizmodo.com/little-florida-memo-could-have-big-impacts-on-flock-cameras-2000805330

WUSF, FDOT revokes permits for Flock cameras on state land. wusf.org/transportation/2026-08-31/fdot-revokes-permits-for-flock-cameras-installed-on-state-land-next-to-roads

Fox 35 Orlando, cameras must come off Florida state roads within 30 days. fox35orlando.com/news/flock-cameras-must-be-pulled-from-florida-state-roads-within-30-days-desantis-orders

Fox Business, DeSantis on highway surveillance. foxbusiness.com/technology/none-governments-damn-business-desantis-pushes-back-highway-surveillance

Florida Phoenix, five sheriffs drop plate readers. floridaphoenix.com/2026/08/31/sheriffs-drop-license-plate-readers-as-desantis-warns-against-surveillance-state/

Fox 4 Dallas, Abbott pauses state funding as Texas cities cancel. fox4news.com/news/abbott-pauses-state-funding-texas-cities-ditch-flock-cameras

Axios, Senate investigation of Flock. axios.com/2026/08/26/flock-cameras-senate-investigation-josh-hawley

Fox News, Hawley on the surveillance network and Texas cancellations. foxnews.com/politics/hawley-rips-flocks-sprawling-surveillance-web-texas-cities-cut-cameras-privacy-data-fears

Route Fifty, five Pennsylvania proposals to regulate plate readers. route-fifty.com/public-safety/2026/08/flock-blockers-5-proposals-seeking-regulate-flock-cameras-and-alprs-pennsylvania/415721/

Arizona Mirror, Attorney General Mayes to examine Flock cameras. azmirror.com/2026/08/27/ag-mayes-to-examine-flock-surveillance-cameras-and-arizonans-privacy-rights/

Futurism, LAPD abandons its Flock contract after false alarm findings. futurism.com/future-society/lapd-abandons-flock-contract-false-alarm

Community Impact, Pflugerville votes unanimously to end its Flock contract. communityimpact.com/pflugerville-hutto/government/pflugerville-city-council-votes-to-end-flock-safety-contract/

KY3, Willard, Missouri chief says the contract will not be renewed. ky3.com/2026/08/21/willard-mo-police-chief-says-flock-safety-contract-will-not-be-renewed/

Cybernews, record contract terminations. cybernews.com/privacy/flock-cameras-us-cities-record-terminations/

The Marshall Project, officers using cameras to track partners. themarshallproject.org/2026/03/07/police-camera-wisconsin-california-colorado

Oaklandside, homeowner groups and powerful surveillance cameras. oaklandside.org/2024/03/27/oakland-homeowner-groups-powerful-surveillance-cameras/

KERA News, Flock skirted regulations as it expanded across north Fort Worth HOAs. keranews.org/government/2024-10-09/surveillance-company-skirted-regulations-as-it-expanded-across-north-fort-worth-hoas

ABC7, Brentwood HOA installs Flock cameras and residents object. abc7news.com/post/neighbors-raise-privacy-concerns-brentwood-hoa-installs-flock-cameras-shadow-lake-housing-development/19666382/

San Francisco Chronicle, the local HOA just captured your license plate, 2018. sfchronicle.com/crime/article/Snap-The-local-HOA-just-captured-your-license-13450561.php

Denverite, Denver council rejects the Flock renewal. denverite.com/2025/05/05/denver-rejects-flock-camera-license-plate-readers/

Denverite, mayor extends the contract at no cost. denverite.com/2025/10/22/mayor-extends-flock-camera-contract-denver/

Denverite, council calls the extension undemocratic. denverite.com/2025/10/29/denver-city-council-opposes-flock-contract-extension/

Denverite, $225,000 settlement for two people detained at gunpoint. denverite.com/2026/05/11/denver-execessive-force-settlement-kansas-shooting/

Gadget Review, cities ditching Flock then hiring Axon for the same job. gadgetreview.com/cities-are-ditching-flock-safety-cameras-then-hiring-axon-to-do-the-same-job

Central Current, Syracuse votes out Flock in favor of a company contracting with ICE. centralcurrent.org/syracuse-lawmakers-ice-out-flock-safety-in-favor-of-company-contracting-with-ice/

Axios San Diego, Flock Nova added without returning to Council. axios.com/local/san-diego/2026/04/29/san-diego-police-flock-nova-surveillance-technology-contract-review

CBS News Colorado, Aurora settles with Brittney Gilliam for $1.9 million. cbsnews.com/colorado/news/aurora-police-settle-colorado-woman-viral-video-held-gunpoint-brittney-gilliam/

KTVU, sheriff pays privacy advocate nearly $50,000 in a plate reader case. ktvu.com/news/sheriff-pays-east-bay-privacy-advocate-nearly-50k-in-license-plate-reader-case

Fox News, how surveillance technology led police to accuse the wrong person, the Bow Mar package case. foxnews.com/tech/how-surveillance-tech-led-police-accuse-wrong-person

9News, more Coloradans stopped after data errors triggered Flock alerts. 9news.com/article/news/local/coloradan-stopped-by-police-data-errors-flock-alerts/73-593a3772-f43d-46b1-8744-af888b9bd7f2

ABC 10News San Diego, two men sue the city over a wrongful arrest linked to Flock data. 10news.com/news/local-news/two-men-sue-the-city-of-san-diego-over-alleged-wrongful-arrest-linked-to-flock-alpr-data

Gibbs Mura, California class action against Flock. classlawgroup.com/flock-license-plate-cameras-face-class-action-lawsuit-by-gibbs-mura-a-law-group

Hoodline, LAPD named in a Flock class action. hoodline.com/2026/05/lapd-dragged-into-court-over-flock-license-plate-dragnet/

WFLX, Flock exposed live police camera feeds to the internet. wflx.com/2026/01/09/flock-safety-exposed-live-police-camera-feeds-internet-data-breach-company-says/

GovTech, Flock Nova and the data controversy. govtech.com/biz/flocks-newest-police-tool-sparks-data-controversy

The Record, Flock adds gunshot and human voice detection. therecord.media/flock-surveillance-technology-gunshot-voice-detection

CNBC, Trump’s Axon purchase and the ICE taser solicitation. cnbc.com/2026/06/29/trump-axon-stock-ice-taser-immigration-enforcement.html

Citizens for Responsibility and Ethics in Washington, the ten companies that added Donald Trump Jr. (advocacy). citizensforethics.org/reports-investigations/crew-reports/companies-see-boost-from-trump-admin-after-adding-don-jr-to-their-board/

Insurance Institute for Highway Safety, red light camera communities. iihs.org/research-areas/red-light-running/red-light-camera-communities

Governors Highway Safety Association, state speed and red light camera laws. ghsa.org/state-laws-issues/speed-red-light-cameras

Department of Justice, former Redflex chief executive sentenced. justice.gov/usao-ndil/pr/former-redflex-ceo-sentenced-30-months-corruption-awarding-chicago-s-red-light-camera

FBI, transportation official sentenced in the red light camera corruption case. fbi.gov/news/stories/transportation-official-sentenced-in-red-light-camera-corruption-case

Department of Justice, former chief executive sentenced in the Ohio red light camera scheme. justice.gov/archives/opa/pr/former-ceo-sentenced-bribery-and-fraud-scheme-involving-red-light-camera-contracts-ohio

TIME, red light camera problems and yellow light timing. time.com/3505994/red-light-camera-problems-tickets/

Chicago Sun-Times, $284.9 million from red light cameras. chicago.suntimes.com/news/emanuel-rakes-in-284-9m-from-red-light-cameras

The Assembly, RedSpeed traffic cameras in North Carolina. theassemblync.com/news/politics/redspeed-traffic-cameras-nc/

Axios Raleigh, Raleigh ends its red light camera program. axios.com/local/raleigh/2024/03/13/raleigh-ready-to-say-goodbye-to-red-light-cameras

Houston Chronicle, voters reject red light cameras. chron.com/cars/article/Houston-voters-reject-red-light-cameras-by-wide-1619320.php

StateScoop, local government cybersecurity staffing and funding, Center for Internet Security data. statescoop.com/local-government-cybersecurity-funding-cis-2025/

Route Fifty, why small municipalities have become cybercriminals’ favorite prey. route-fifty.com/cybersecurity/2026/03/why-small-municipalities-have-become-cybercriminals-favorite-prey/412364/

E8. United Kingdom and European sources

National ANPR Standards for Policing and Law Enforcement, version 3.5, July 2026. assets.publishing.service.gov.uk/media/6a69ab2e16bc92f51e1a42ae/NASPLE_v3.5.pdf

National ANPR Service Data Protection Impact Assessment, updated April 2026. gov.uk/government/publications/national-anpr-service-data-protection-impact-assessment

Protection of Freedoms Act 2012, Part 2 Chapter 1, the surveillance camera code. legislation.gov.uk/ukpga/2012/9/part/2/chapter/1

Biometrics and Surveillance Camera Commissioner, letter to the Secretary of State for Transport, October 2023. assets.publishing.service.gov.uk/media/652665d3aea2d00013219b5f/Letter_to_Secretary_of_State_for_Transport_FS.pdf

Surveillance Camera Commissioner speech, The long and winding road, November 2017. gov.uk/government/speeches/the-long-and-winding-road-the-humble-number-plate

Biometrics and Surveillance Camera Commissioner, National ANPR Conference speech, 2021. gov.uk/government/news/biometrics-and-surveillance-camera-commissioner-speech-at-the-national-anpr-conference-2021

House of Lords Justice and Home Affairs Committee, Technology rules, HL Paper 180, March 2022. publications.parliament.uk/pa/ld5802/ldselect/ldjusthom/180/18002.htm

European Digital Rights, the Information Commissioner ruling on the Royston ANPR cordon. edri.org/our-work/edrigramnumber11-15uk-vehicle-recognition-system-ruled-illegal/

General Data Protection Regulation, Regulation (EU) 2016/679. eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32016R0679

Law Enforcement Directive, Directive (EU) 2016/680. eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32016L0680

German Federal Constitutional Court, automated plate recognition, March 2008. bundesverfassungsgericht.de/SharedDocs/Pressemitteilungen/EN/2008/bvg08-027.html

German Federal Constitutional Court, Bavarian Police Act, decided December 2018. bundesverfassungsgericht.de/SharedDocs/Pressemitteilungen/EN/2019/bvg19-008.html

German Federal Constitutional Court, Baden-Wurttemberg and Hesse, decided December 2018. bundesverfassungsgericht.de/SharedDocs/Pressemitteilungen/EN/2019/bvg19-009.html

Digital Rights Ireland, Joined Cases C-293/12 and C-594/12. eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:62012CJ0293

Tele2 Sverige and Watson, Joined Cases C-203/15 and C-698/15. eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:62015CJ0203

La Quadrature du Net and Others, Court of Justice press release, October 2020. curia.europa.eu/jcms/upload/docs/application/pdf/2020-10/cp200123en.pdf

European Court of Human Rights, personal data protection factsheet, including Catt v. United Kingdom. echr.coe.int/documents/d/echr/fs_data_eng

European Digital Rights, Denmark’s targeted ANPR retention turned into mass surveillance. edri.org/our-work/denmark-targeted-anpr-data-retention-turned-into-mass-surveillance/

Washington Revised Code 46.63.220, automated traffic safety cameras. app.leg.wa.gov/RCW/default.aspx?cite=46.63.220

California Vehicle Code 21455.5, automated enforcement systems. law.justia.com/codes/california/code-veh/division-11/chapter-2/article-3/section-21455-5/

California Civil Code 1798.90.55, the public comment requirement before operating an ALPR system. law.justia.com/codes/california/code-civ/division-3/part-4/title-1-81-23/section-1798-90-55/

Seattle Municipal Code Chapter 14.18, acquisition of surveillance technology. library.municode.com/wa/seattle/codes/municipal_code/349918?nodeId=TIT14HURI_CH14.18ACUSSUTE_14.18.050EQIMAS

E9. Apex and Wake County specific sources

Town of Apex, cybersecurity incident information page. apexnc.org/1983/Cybersecurity-Incident-Information

The Peak Weekly, the Apex Flock contract, term, consent agenda vote and equipment. newsletter.thepeakweekly.com/p/tinkle-bells-flock-cameras

WRAL, Apex council reviews its Flock cameras, August 2026. wral.com/news/local/apex-flock-cameras-meeting-august-2026/

Hoodline, Apex residents pack a council meeting over Flock cameras and federal data sharing. hoodline.com/2026/08/apex-residents-pack-council-meeting-over-flock-cameras-federal-data-sharing-fears/

Indy Week, what the Flock, an explainer on the cameras deployed across the Triangle. indyweek.com/news/what-the-flock-an-explainer-on-the-controversial-license-plate-reading-cameras-deployed-across-the-triangle/

CBS17, Apex cameras continue to raise concern as an Axon contract renewal is on the table. cbs17.com/news/local-news/wake-county-news/cameras-in-apex-continue-to-raise-concern-as-an-axon-contract-renewal-is-on-the-table/

Comparitech, Town of Apex warns 22,000 people of a two year old data breach. comparitech.com/news/town-of-apex-nc-warns-22000-people-of-2-year-old-data-breach/

DataBreaches.net, Apex recovers stolen personal data after 22,000 impacted. databreaches.net/2026/04/03/apex-recovers-stolen-personal-data-after-about-22k-impacted-in-cyberattack-town/

WRAL, Apex utilities resume late fees after the cyber attack, February 2026. wral.com/news/local/apex-utilities-resume-late-fees-cyber-attack-feb-2026/

WRAL, Apex works through billing issues months after the cyber attack, January 2025. wral.com/news/local/apex-works-through-billing-issues-months-after-cyber-attack-jan-2025/

ABC11, Apex customers hit with high water bills after the attempted ransomware attack. abc11.com/post/apex-cyberattack-customers-high-water-bills-attempted-ransomware-mayor/15428577/

ABC11, Apex hires a firm to investigate the billing errors. abc11.com/post/cyberattack-high-utility-bills-apex-investigated-city-hired-firm/15620309/

CBS17, Apex recovers stolen personal data. cbs17.com/news/local-news/wake-county-news/apex-recovers-stolen-personal-data-after-about-22k-impacted-in-cyberattack-town/

WRAL, Raleigh plate reader cameras help police make 41 arrests in six months. wral.com/story/license-plate-reading-cameras-help-raleigh-police-make-41-arrests-in-6-months/20702950/

WRAL, Wake County woman charged in a child abduction located with Flock cameras. wral.com/news/local/wake-county-woman-flock-cameras-child-abduction-july-2026/

North Carolina League of Municipalities, Property and Casualty Trust. nclm.org/insurance-risk-management/property-casualty-trust/

North Carolina League of Municipalities law enforcement risk review, Fuquay-Varina Police Department, 2023. fuquay-varina.org/DocumentCenter/View/8782/Fuquay-Varina-Police-Department-Risk-Review-2023

NeighborhoodScout, Apex crime data derived from FBI reporting (secondary aggregator). neighborhoodscout.com/nc/apex/crime

DeFlock Apex (advocacy). deflockapex.org/

DeFlock Apex, draft ALPR governance ordinance for the Town of Apex, June 2026 (advocacy, not enacted). deflockapex.org/docs/apex-alpr-governance-ordinance.pdf

Chatham Journal, Pittsboro’s Flock fight, including per camera pricing. chathamjournal.com/2026/05/22/pittsboros-flock-fight-how-license-plate-cameras-became-a-test-of-public-safety-privacy-and-local-control/

Sandhills News, Moore County residents told no vote or referendum is required. sandhills.news/2026/08/18/moore-county-flock-cameras/

NCLocal, western North Carolina counties weigh surveillance. nclocal.org/2026/07/30/balancing-public-safety-and-privacy-wnc-wrestles-with-surveillance-issues/

E10. Contract and cancellation trackers, all advocacy compiled

Who Approved This, Flock contract and approval method tracker. whoapprovedthis.org/

DeFlock Wilmington, North Carolina plate reader law and records repository. deflockilm.org/north-carolina-license-plate-reader-law/

Flip the Flock, state by state legislative tracker. fliptheflock.org/states/north-carolina/

Finding Flock, contract cancellation tracker. findingflock.com/learn/flock-contract-cancellations

Newsweek, map of cities that have rejected or deactivated Flock cameras. newsweek.com/map-cities-rejected-deactivated-flock-cameras-12253499

Have I Been Flocked, contract document repository. haveibeenflocked.com/

E11. Town of Apex public records, and the sources behind Section 6

Adam Walls (Compliance Specialist, APD) to Chief Ryan Johansen, “ALPR Notes,” April 9, 2026, attaching the spreadsheet Flock Success (11 named incidents; 91 CAD calls since September 1, 2025). Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Flock account executive to Deputy Chief Trevor Materasso, “Re: Deployment,” April 10 to 16, 2026 (the disputed count of 16 active deployments). Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Deputy Chief Trevor Materasso and staff, “RE: FLOCK Amendment,” March 10 to April 1, 2026 (three Flex cameras converted to fixed; $650 per camera installation). Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Apex PD staff to Flock, “Re: Account Update – FLOCK,” April 7 to 17, 2026 (the addendum “Adding additional 10 cameras to existing contract”; purchase order deadline). Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Erika Sacco (Information Technology) to Chief Ryan Johansen, “RE: Flock ALPR follow-up,” May 1, 2026, replying to the Chief of April 14, 2026 (the unresolved reported camera compromise). Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Jennifer Conley (Criminal Intelligence Analyst, APD) to Deputy Chief Trevor Materasso, “State Agency Requesting Flock Access,” April 8 to 17, 2026 (two federal agencies requesting access on April 15). Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Flock Safety to Apex NC PD, “MacGregor Downs Country Club (NC) shared Flock cameras with you,” April 3, 2026. Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Town of Apex, Public Records Request PUPR-2026-092, internal routing thread, February 17 to April 2, 2026. Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Competing LPR vendor to Deputy Chief Trevor Materasso, “Re: Evaluating LPR vendors?” March 3 to April 10, 2026. Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Resident letter to Mayor Jacques Gilbert, “Concern Over the Use of Flock Cameras,” April 16, 2026. Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Resident memorandum, “Flock Concerns,” July 2026 (advocacy; individually footnoted, reproduced in Section 6.6 as resident-compiled). Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

Resident memorandum, “Flock Outside Access,” August 2026 (advocacy; the 994 organizations figure, the SB 153 agencies, the misuse tally). Town of Apex, released under N.C.G.S. Chapter 132. Copy on file with the requester

IPVM, Flock ALPR accuracy evaluation and Flock’s response, cited by the July 2026 memo for the 10 percent misread figure. ipvm.com/reports/flock-ipvm-state

Senator Ron Wyden, letter to the Federal Trade Commission on Flock’s cybersecurity practices. wyden.senate.gov/imo/media/doc/wyden_letter_to_ftc_on_flockpdf.pdf

University of Washington, Leaving the Door Wide Open, on Border Patrol searches in Washington police network audits (2025). jsis.washington.edu/humanrights/2025/10/21/leaving-the-door-wide-open/

Footnote 4A, on Flock’s February 2026 terms and conditions changes (advocacy; includes a marked-up comparison). footnote4a.org/news/terms-feb2026

Footnote 4A, on the December 2025 reduction of Network Audit fields (advocacy; includes Flock executive emails). footnote4a.org/news/secret-searches-part2

Footnote 4A, on Flock order forms and linked terms (advocacy). footnote4a.org/news/trojan-contracts

North Carolina Department of Transportation, C540 Phase 2 RFP, cited for the 99.9 percent toll imaging accuracy requirement. connect.ncdot.gov/business/Turnpike/ProcurementsLibrary/C540%20Phase%202%20RFP_1.0.pdf

E12. Axon Enterprise: filings, court records, ownership and governance

Axon Enterprise, Annual Report on Form 10-K for FY2025 (vendor filing; product lines, competitors, segment revenue). sec.gov/Archives/edgar/data/1069183/000162828026011360/axon-20251231.htm

Axon Enterprise, Annual Report on Form 10-K for FY2024 (vendor filing; the only mention of Trump is a tariff risk factor). sec.gov/Archives/edgar/data/1069183/000106918325000019/axon-20241231.htm

Axon Enterprise, 2026 proxy statement, Form DEF 14A (shares outstanding, board of directors, beneficial ownership table at pages 29 to 31). sec.gov/Archives/edgar/data/1069183/000162828026025521/axon-20260416.htm

Axon Q4 and full year 2025 results (vendor). investor.axon.com/2026-02-24-Axon-reports-Q4-2025-revenue-of-797-million,-up-39-year-over-year

Axon Q4 2025 shareholder letter (vendor; the ALPR entry claim quoted in Section 11.2). filecache.investorroom.com/mr5ir_axon/590/AXON_Q4_2025_ShareholderLetter.pdf

Axon Q2 2026 results (vendor). prnewswire.com/news-releases/axon-reports-q2-2026-revenue-of-904-million-up-35-year-over-year-302844105.html

Axon Body 4 launch (vendor). investor.axon.com/2023-04-11-Axon-launches-next-generation-body-camera-with-more-features-to-never-miss-a-moment

Axon, more than 2,000 law enforcement agencies figure (vendor, January 2024). investor.axon.com/2024-01-25-Axon-Launches-Axon-Body-Workforce,-New-Body-Camera-To-Protect-Frontline-Workers

Axon Air and Drone as First Responder (vendor product pages). axon.com/dfr

Axon and Skydio drone partnership (vendor, June 2024). investor.axon.com/2024-06-20-Axon-and-Skydio-partner-to-deliver-scalable-drone-offering-for-public-safety,-including-Drone-as-First-Responder-solution

Axon to acquire Dedrone, counter-drone (vendor, May 2024). investor.axon.com/2024-05-06-Axon-to-acquire-Dedrone,-accelerating-the-next-generation-of-drone-solutions-to-protect-more-lives-in-more-places

Sky-Hero joins Axon, indoor tactical drones and ground robots (vendor, July 2023). axon.com/blog/our-next-step-in-robotics-sky-hero-joins-forces-with-axon

Axon announces TASER drone development for mass shootings, including schools (vendor, June 2, 2022). investor.axon.com/2022-06-02-Axon-Announces-TASER-Drone-Development-to-Address-Mass-Shootings

NPR, Axon halts plans for Taser drone as nine on ethics board resign (June 6, 2022). npr.org/2022/06/06/1103285030/axon-halts-plans-for-taser-drone-as-9-on-ethics-board-resign

The Record, Axon still wants to put Taser drones in your kids school (January 2023). therecord.media/exclusive-axon-still-wants-to-put-taser-drones-in-your-kids-school

Axon unveils TASER 10 (vendor, January 2023). investor.axon.com/2023-01-24-Axon-Unveils-TASER-10

Axon Fleet 3 with ALPR now shipping (vendor, June 2021; the release in which Axon resold Flock’s fixed ALPR). investor.axon.com/2021-06-30-Axon-Fleet-3-Has-Arrived-Next-Generation-In-Car-Video-System-with-ALPR-Now-Shipping-to-Public-Safety-Agencies

Axon announces fixed ALPR, Outpost and Lightpost (vendor, April 22, 2025). axon.com/newsroom/press-releases/axon-announces-new-fixed-ALPR-camera-solutions-and-next-gen-AI-advancements-to-expand-real-time-public-safety-ecosystem

Axon Vehicle Intelligence, including the invite-based sharing claim (vendor product page). axon.com/products/vehicle-intelligence

Axon acquires Fusus, real-time crime centers (vendor, February 2024). investor.axon.com/2024-02-01-Axon-Accelerates-Real-Time-Operations-Solution-with-Strategic-Acquisition-of-Fusus

Axon Fusus product page, including Community Connect for private cameras (vendor). axon.com/products/axon-fusus

Axon and Ring community request partnership (vendor, October 2025). axon.com/blog/building-safer-communities-together-axon-and-ring

Axon launches Draft One AI report writing (vendor, April 2024). investor.axon.com/2024-04-23-Axon-reimagines-report-writing-with-Draft-One,-a-first-of-its-kind-AI-powered-force-multiplier-for-public-safety

Electronic Frontier Foundation, Axon’s Draft One is designed to defy transparency (advocacy, July 2025; the retained-draft finding). eff.org/deeplinks/2025/07/axons-draft-one-designed-defy-transparency

Axon to acquire Prepared, AI 911 call triage (vendor, September 2025). investor.axon.com/2025-09-23-Axon-to-acquire-AI-powered-911-communications-company-Prepared,-strengthening-the-public-safety-ecosystem-from-call-to-closure

Axon to acquire Carbyne, next-generation 911 (vendor, November 2025). investor.axon.com/2025-11-04-Axon-to-Acquire-Carbyne,-Uniting-Cloud-Infrastructure-and-AI-to-Redefine-the-911-Experience

Axon, 4 Reasons Why Agencies Have Opted to Sole Source (vendor; the procurement marketing quoted in Section 11.3). axon.com/news/4-reasons-why-agencies-have-opted-to-sole-source

OMNIA Partners cooperative purchasing, Axon supplier page. omniapartners.com/suppliers/axon

Statement of resigning Axon AI Ethics Board members, hosted by the Policing Project at NYU School of Law (primary). policingproject.org/statement-of-resigning-axon-ai-ethics-board-members

Federal Trade Commission challenges Axon’s consummated acquisition of Vievu (January 3, 2020). ftc.gov/news-events/press-releases/2020/01/ftc-challenges-consummated-merger-companies-market-body-worn

Federal Trade Commission case docket, Axon Enterprise and Safariland. ftc.gov/legal-library/browse/cases-proceedings/1810162-axon-enterprise-safariland-matter

Axon announces FTC dismissal of the antitrust complaint (vendor, October 2023). investor.axon.com/2023-10-09-Federal-Trade-Commission-Dismisses-Antitrust-Complaint-Against-Axon

Axon Enterprise, Inc. v. FTC, 598 U.S. 175 (2023), slip opinion (the holding is procedural). supremecourt.gov/opinions/22pdf/21-86_l5gm.pdf

In re Axon Enterprise, Inc. Vievu Antitrust Litigation, D.N.J. No. 3:23-cv-07182, counsel case page. cohenmilstein.com/case-study/in-re-axon-vievu-antitrust-litigation/

Township of Howell v. Axon Enterprise, complaint as filed (the lock-in and pricing allegations quoted in Section 11.3). classaction.org/media/township-of-howell-monmouth-county-new-jersey-v-axon-enterprise-inc-et-al.pdf

Axon SEC filing characterizing the municipal antitrust suit as based largely on unproven allegations (vendor). sec.gov/Archives/edgar/data/1069183/000106918325000181/R20.htm

BlackRock, Inc., Schedule 13G on Axon Enterprise, 9.5 percent as of December 31, 2025. sec.gov/Archives/edgar/data/1069183/000201238326000646/primary_doc.xml

The Vanguard Group, Schedule 13G amendment on Axon Enterprise, 11.86 percent as of September 30, 2025. sec.gov/Archives/edgar/data/1069183/000010290925000037/primary_doc.xml

The Vanguard Group, Schedule 13G amendment reporting zero shares following an internal reorganization effective January 12, 2026 (not a divestment). sec.gov/Archives/edgar/data/1069183/000010290926000719/primary_doc.xml

Patrick W. Smith, Form 4, June 2026 (3,060,997 shares held directly; sales under a Rule 10b5-1 plan). sec.gov/Archives/edgar/data/0001069183/000162828026041656/xslF345X06/wk-form4_1780952044.xml

Donald J. Trump, Office of Government Ethics Form 278-T periodic transaction report, filed May 2026 (the Axon purchase line, primary document). static.notus.org/2b/18/a1761d44459fb1c67e196c1b3629/trump-donald-j-05-08-2026-278t2.pdf

Notus, Trump bought corporations’ stock as his administration boosted their business. notus.org/money/donald-trump-stock-investments-palantir-axom-nvidia

CNBC, Trump bought as much as $5 million in Axon stock before ICE sought a $220 million Taser deal (June 2026). cnbc.com/2026/06/29/trump-axon-stock-ice-taser-immigration-enforcement.html

Scripps News, Trump’s investment accounts show stock trades in companies awarded immigration contracts (the $1.26 million across 24 trades figure). scrippsnews.com/investigations/ice-inc/trumps-investment-accounts-show-stock-trades-in-companies-awarded-immigration-contracts

1789 Capital, Donald Trump Jr. partner page (his actual role; Axon is not in the portfolio). 1789capital.vc/don-trump-jr

Citizens for Responsibility and Ethics in Washington, companies that added Donald Trump Jr. to a board (advocacy; Axon is not among them). citizensforethics.org/reports-investigations/crew-reports/companies-see-boost-from-trump-admin-after-adding-don-jr-to-their-board/

Unusual Machines announces Donald Trump Jr. to its advisory board, SEC exhibit (the drone company most likely confused with Axon). sec.gov/Archives/edgar/data/1956955/000168316824008412/umac_ex9901.htm

United States Senate Lobbying Disclosure Act filings, Axon 2025. lda.gov/api/v1/filings/?client_name=Axon&filing_year=2025

OpenSecrets, Palantir, Axon and Parsons triple lobbying expenditures while winning ICE contracts (2026). opensecrets.org/news/2026/04/palantir-axon-parsons-triple-lobbying-expenditures-while-raking-in-millions-from-ice-contracts/

Federal Election Commission, Axon Enterprise Inc Political Action Committee, committee C00890590. fec.gov/data/committee/C00890590/

ExecutiveGov, Ronald Vitiello named Acting Deputy Commissioner of Customs and Border Protection (the revolving door in Section 11.7). executivegov.com/articles/ronald-vitiello-named-deputy-commissioner-cbp

Thomas Brewster, Forbes, Axon Says AI Police Reports Save Time. Public Records Show They Get Facts Wrong (July 22, 2026). The public records investigation underlying Section 11.9: the Lafayette, Indiana Form One emails, the Manchester and Anchorage cancellations, the King County prosecutor refusal, and the Richmond, California AI-generated reports. forbes.com/sites/thomasbrewster/2026/07/22/axon-says-ai-police-reports-save-time-public-records-show-they-get-facts-wrong/

Ian Adams et al., study of AI-assisted police report writing at the Manchester, New Hampshire Police Department, Journal of Experimental Criminology (2024). The only peer-reviewed test of the time-savings claim; cited here as reported by Forbes above, and worth obtaining in full before the meeting. link.springer.com/journal/11292

Oregon Public Broadcasting, Flock competitors seek to benefit from public backlash (August 2026). opb.org/article/2026/08/22/flock-competitors-seek-to-benefit-from-public-backlash/

App E13The abuse database, record by record

All 194 records captured from the Institute for Justice database on September 3, 2026, in the order the database presents them, newest first. Each line gives the date, the location, the category and the underlying source the database cites. Three of the 197 records the database displayed, all from 2021 and 2022, could not be read in full and are not listed. Protocol prefixes are omitted from the links.