Last reviewed: October 20, 2026
The risks of searchable surveillance databases are not purely theoretical.
Real systems have produced real misuse, incorrect matches, access-control problems and frightening encounters for innocent people.
That does not mean every officer abuses these systems.
It means four things are worth remembering:
- Someone can abuse access.
- Technology can get it wrong.
- Data can spread farther than expected.
- An innocent person can pay the price.
We keep categories separate.
Not every troubling incident is the same thing.
We distinguish:
- Personal misuse - authorized access allegedly or officially used for an improper personal purpose.
- False matches / bad data - a system or related database produces incorrect information that contributes to action against the wrong person or vehicle.
- Interstate or sensitive access - surveillance data is used across jurisdictional lines or for especially sensitive investigations.
- Vendor or configuration failure - access settings, permissions or product behavior do not match what an agency or law intended.
- Security - technical weaknesses or vulnerabilities affecting the system.
- Investigative expansion - newer tools materially expand what can be searched or inferred.
We also label status carefully: confirmed fact, charged/formally accused, official finding, credibly reported, pending litigation, or unresolved.
Personal misuse: former romantic partners and other improper searches
Public reporting and government records have documented law-enforcement database misuse involving current or former romantic partners.
Recent ALPR-specific reporting has included cases in which investigators or officers were accused of using Flock or related databases to search vehicles connected to former partners without a legitimate investigative purpose.
These cases matter because they illustrate a basic access-control problem:
A legitimate credential can still be misused by the person who holds it.
Audit logs may help expose that later.
They do not prevent the search from happening in the first place.
Albany County, New York
In 2026, public reporting described an Albany County investigator who was criminally charged after allegedly conducting thousands of unauthorized searches, including a large number involving a vehicle associated with an ex-girlfriend.
Status: Charged / formally accused.
The allegation is serious.
It should be described as a charge/allegation unless and until the legal process establishes more.
Times Union reporting on the Albany County charges ↗
The larger lesson is about privileged access: powerful databases create opportunity for personal misuse even where most users follow the rules.
Alpharetta, Georgia
Public-record reporting by WIRED described an internal investigation involving an Alpharetta officer’s searches connected to a former romantic partner and another officer.
The officer resigned.
Status: Internal investigation / policy finding reported from public records.
WIRED: A Georgia Cop Used Flock to Track 2 Other Cops ↗
Again, the point is not that every officer behaves this way.
The point is that surveillance systems must be evaluated partly by what becomes possible when access is abused.
Eastpointe, Michigan: two employee-misuse findings and termination
On September 16, 2026, Eastpointe Police Chief Corey Haines said an internal investigation found that two Police Department employees misused Flock. According to the chief, one employee used the system for a personal purpose and received severe discipline. A second employee committed a policy violation, with that matter still under investigation at the time of the announcement.
Status: Official findings described by the police chief; one related matter remained under investigation.
The preceding night, City Council voted 4-1 to terminate the Flock agreement. The department discontinued using Flock while the city worked through termination. Reporting said physical removal had not yet been scheduled.
CBS Detroit: police chief describes the two misuse findings ↗
WXYZ: termination vote, discontinued use and removal still pending ↗
This is a particularly relevant local case. Auditability and discipline are better than no oversight, but they responded after employees already had access and misuse occurred. The council then exercised its authority to stop continued use.
Read the full Eastpointe dossier →
October 2026 brought another cluster of alleged and confirmed misuse.
These cases are separate matters with different procedural statuses. They should not be blended into one claim, but together they test the argument that credentials, required search reasons and after-the-fact audits are enough to prevent abuse.
Savannah, Georgia: a credential was shared with an ICE agent
Savannah Police fired Officer Amber Harpe after an internal investigation found she gave her Flock login to her husband, an Immigration and Customs Enforcement agent who did not have his own Flock account through that federal job. Department records reviewed by WJCL tied 983 searches involving 97 identifiable plates across six states to her account between January and July 2026.
The investigation found unauthorized access and a policy breach. It does not establish that Harpe personally ran every search, because she said her husband ran most of them. ICE confirmed it was investigating its agent’s involvement.
Status: Officer fired after internal finding; federal involvement under investigation.
WJCL: department records and interview audio ↗
WTOC: Savannah and ICE confirm the access issue ↗
Athens-Clarke County, Georgia: searches of an ex-girlfriend’s vehicle
Athens-Clarke County Police fired Officer Tyler Vaughn after an internal investigation found that he repeatedly searched a plate associated with a former girlfriend over about a week in 2025. The department requested an independent Georgia Bureau of Investigation review.
Status: Fired after internal finding; independent investigation requested.
CBS Atlanta: department finding and investigation status ↗
Sharonville, Ohio: a lieutenant was indicted
The Hamilton County Prosecutor announced that a grand jury indicted a Sharonville police lieutenant on a felony unauthorized-use charge. Prosecutors allege he searched Flock 29 times between December 2024 and April 2026 to locate someone he knew for personal reasons unrelated to police work.
Status: Criminal charge pending; an indictment is not a conviction.
Hamilton County Prosecutor’s official announcement ↗
Baldwin County, Georgia: two officers charged
Booking records show two law-enforcement officers were charged October 9 with violation of oath and illegal use or obtainment of law-enforcement-captured plate data. One employing agency described its former deputy’s case as involving Flock. Both men lost their jobs.
The public booking reports do not explain what data was allegedly sought or why. The cases remain pending.
Status: Criminal charges pending; the records do not establish guilt.
41NBC: booking records, charges and evidence limits ↗
The narrow conclusion is strong enough: required credentials and logged searches did not prevent the alleged conduct from occurring. Audits and discipline may expose misuse later; they cannot make a completed search disappear.
Sensitive interstate searches
In 2025, reporting and official responses drew attention to an abortion-related investigation in which a nationwide ALPR search touched a very large network of cameras, including concerns about access to Illinois data despite state restrictions on certain uses involving abortion, immigration and gender-affirming care.
Subsequent scrutiny in Illinois identified broader access-control concerns and led to changes affecting which agencies could access Illinois-linked data.
Status: Official state finding / response.
Illinois Secretary of State audit announcement ↗
The importance is not one political issue.
It is that a locally installed camera can become relevant to an investigation initiated somewhere else.
That is why sharing configuration matters.
Cleveland schools: public description and actual access did not match.
The Cleveland Metropolitan School District operates about 200 ALPRs at 62 schools. In June 2026, the district told Signal Cleveland that only district security personnel and Cleveland police could access its data. An October board presentation later disclosed that more than 500 outside agencies had access and conducted more than 130,000 searches in June, and that the district did not revoke outside access until July 6.
An internal audit later found no significant deficiencies after the district changed the settings. That does not erase the earlier discrepancy. It shows why residents need configuration exports and audit records, not only a verbal description of who supposedly has access.
Status: District presentation and audit process reported; outside access later disabled.
Axios Cleveland: school-camera access and renewal review ↗
Ventura, California
In 2026, Ventura Police publicly disclosed that a vendor configuration had allowed two out-of-state agencies to run nationwide queries despite intended restrictions.
Status: Confirmed by the local agency.
Ventura Police Department explanation ↗
Configuration and policy are not the same thing.
A written rule does not matter if the technical permissions do not enforce it or if people do not understand how the system is actually configured.
False plate reads can become dangerous quickly.
ALPR results are investigative information, not magical truth.
Public reporting has documented innocent motorists being stopped after incorrect plate information or misreads contributed to police believing a vehicle was associated with a crime or stolen car.
One 2026 case in Sherwood, Arkansas involved an incorrect plate-character read contributing to an innocent couple being detained at gunpoint while their six-week-old baby was in the vehicle.
Status: Documented technical error / wrongful stop.
Institute for Justice compilation and source record ↗
The specific facts of any such case should be sourced carefully.
The general lesson is simple:
Verify the machine before acting as though the machine is certain.
Bad data can come from more than the camera.
Even a technically correct camera read can become dangerous when another database contains stale or incorrect information.
ALPR systems interact with larger investigative ecosystems.
Potential failure points include:
- camera recognition error;
- incorrect hot-list data;
- stale stolen-vehicle records;
- wrong database associations;
- mistaken interpretation by the user;
- and failure to verify before a high-risk stop.
A precise-looking screen does not eliminate human judgment.
Broader law-enforcement database misuse matters too.
ALPRs are not the only searchable government databases that have been misused.
Recent investigations have documented numerous officers accused of searching law-enforcement databases for personal reasons, including information connected to romantic partners. In August 2026, The Washington Post reported at least 50 officers charged with or accused of misusing Flock or other license-plate-reader systems, and later reporting increased the identified total.
Washington Post investigation ↗
That broader history matters because the risk does not come from the letters “ALPR.”
It comes from giving people privileged access to sensitive data.
Security concerns need careful wording.
Security researchers have identified hardware, software and deployment weaknesses involving surveillance technology.
Those findings deserve attention.
A vulnerability report is not proof that customer data was stolen. We keep those claims separate.
As of our latest review, we have not established a confirmed breach of Flock’s central cloud platform involving customer-data exfiltration. If strong evidence changes that conclusion, we will update the claim.
That narrow statement is not the same as saying no meaningful security incident has occurred.
On December 22, 2025, 404 Media reported finding at least 60 internet-exposed Condor interfaces that did not require a login, allowing public access to live video, roughly 30 days of archives and, in some instances, settings, logs or other administrative functions. The views included children on a playground and adults on paths and in parking areas; cameras were observed automatically zooming toward or following people. A reporter visited one location and verified that colleagues could watch him through the exposed interface.
The interfaces were found through Shodan, a public search service for internet-connected devices. Flock described the incident as a limited carrier-testing configuration affecting a small number of cameras, said it corrected the problem and said it deployed additional safeguards.
No reported abduction resulted, and this is not proof that every Condor camera was exposed. The documented exposure nevertheless could have allowed a stranger to learn when a child or adult appeared alone, their route or a repeated routine. That is a concrete stalking and child-safety risk created by the access, not an outcome we claim occurred.
404 Media: exposed Condor feeds and archives ↗
WPTV: Flock confirms and responds to the exposed feeds ↗
Separate laboratory research by Jon “GainSec” Gaines documented 51 findings involving Flock hardware, embedded software and apps, including 22 assigned CVEs and eight awaiting assignment when the paper was published. The researcher says the testing used lawfully acquired equipment in a controlled lab and did not access production systems or customer data.
Flock acknowledged the research and the assigned vulnerability identifiers. It said the findings generally required physical access and specialized hardware knowledge, that it had no indication customers were exploited and that it was deploying updates or mitigations. A vulnerability is not proof of exploitation; it is evidence that the risk should be evaluated rather than dismissed.
Primary research repository and disclosure notes ↗
Archived research paper on Zenodo ↗
Flock’s response to the compiled device research ↗
SimeonOnSecurity: accessible secondary overview ↗
WIRED later reported that people who physically removed and examined a Flock camera extracted device material and imagery from it. Unauthorized removal or access is illegal and contrary to this campaign’s rules. The technical finding still matters when evaluating physical security and claims about what a person with device access could retrieve.
WIRED: reported data extraction from a physically removed Flock camera ↗
Those are different event types from a central-cloud breach. Precision lets residents reject both exaggeration and an overly broad “never hacked, therefore no security problem” response.
Security criticism should be as precise as every other criticism.
Protest-related searches have been documented.
In 2026, the Electronic Frontier Foundation analyzed audit material involving protest-related Flock searches. The Record reported that more than 50 agencies had queries associated with demonstrations including No Kings, 50501 and Hands Off events.
Status: Public-record analysis / credible reporting.
The records do not establish that every search was unlawful or that every agency used the same method. They do establish that searching in connection with political demonstrations is not merely a hypothetical use.
The Record: EFF analysis of protest-related Flock searches ↗
New AI and investigative features expand the consequences of access.
The more a surveillance platform can search, connect and infer, the more consequential both legitimate use and misuse become.
Searches by vehicle description, natural-language queries, shared networks, convoy relationships and multi-location patterns can be useful to investigators.
They also make unauthorized or mistaken use potentially more revealing.
Capability does not prove misuse.
Capability changes what misuse could expose.
“There are audit logs” is not the whole answer.
Audit logs are good.
Use them.
They can deter some misuse and help uncover other misuse.
But an audit log is retrospective.
It can tell us someone searched for an ex-partner after the search occurred.
It cannot un-search the database.
“Our officers would never do that” is not a security model.
Most users of most systems probably behave appropriately.
Security is designed around the reality that some people will not.
That is why banks use permissions.
Hospitals audit medical records.
Businesses limit access to customer information.
And governments should evaluate surveillance infrastructure based on what happens when procedures fail as well as when they work perfectly.
We do not ask you to take our word for it.
Each public incident we rely on should ultimately point back to strong supporting evidence: court records, official findings, audit material, government statements or credible reporting grounded in records.
We will not deliberately inflate one allegation into proof that an entire profession behaves the same way.
We also will not pretend documented misuse disappears because most users behave appropriately.
The policy question remains.
Every powerful system creates some possibility of misuse or error.
That does not automatically mean every powerful system must be prohibited.
The question is whether the benefit justifies creating the capability in the first place.
With mass vehicle-tracking infrastructure, our answer is no.
The safest way to prevent misuse of a database of ordinary people’s movements is not to build that database.
Compare Flock claims, documented incidents and common objections →