The strongest case against mass surveillance does not depend on saying that every camera does everything, every user acts improperly or every safeguard is fake.
It depends on putting the whole record in view.
Flock says its systems help solve crimes, use access controls, create audit logs, limit retention and do not use facial recognition. Those claims deserve to be represented accurately.
The public record also shows that:
- Flock’s broader platform can search enabled video for people by clothing and other visible characteristics;
- company personnel used private camera feeds at sensitive locations in sales demonstrations;
- Flock-controlled demonstration accounts searched real customer cameras while testing sensitive natural-language queries;
- some Condor live feeds and archives were exposed to the public internet without a login;
- authorized users have been accused of searching for former partners;
- searches have crossed state lines for abortion, immigration and protest-related investigations;
- configuration errors have defeated intended access limits; and
- information selected, downloaded or placed in a case can remain after routine deletion.
No one incident proves every possible criticism. Together, they make one thing difficult to dismiss: the risk is built from the collection, the searchable platform, the permissions and the connections—not merely from the word “plate.”
How to read this page
We keep several categories separate:
- Documented event means a government record, company statement or strong records-based report establishes that something happened.
- Vendor capability means Flock advertises or documents a feature. It does not prove a particular municipality bought or used it.
- Allegation means a person has been accused or charged; it is not treated as a final finding.
- First-person account means someone publicly described an experience that has not necessarily been independently verified.
- Policy inference means we explain what the evidence means for a public decision without pretending the evidence proves more than it does.
This is a curated, living evidence file—not a claim that we have found every incident.
“It only records a license plate and a car.”
That is an incomplete description of the current platform.
A standard Falcon roadside LPR is vehicle-focused. Flock says people-search features do not operate on those plate-reader cameras, and we have not found evidence that the standard Falcon product performs facial recognition.
But Flock’s own FreeForm material says authorized users can search enabled video using descriptions such as a person’s clothing or accessories. Its video products include person alerts, recorded video, remote pan/tilt/zoom and digital or optical zoom. Its platform can place video, vehicle observations and shared sources into a common workflow.
Flock video-camera product page ↗
Flock and its critics sometimes use the word track differently. A company representative told InvestigateTV that individual sightings are not continuous tracking. Flock training material has also described following vehicles “from location to location” and tracking a suspect’s movements, while Condor video can follow a person in view.
The plain-language point is more important than the label: searchable observations from multiple places can reveal movement, whether the vendor calls that tracking, point sightings or an investigative timeline.
InvestigateTV: Flock’s public wording compared with its training materials ↗
Sensitive private-camera feeds were used in sales demonstrations.
In Dunwoody, Georgia, public records and reporting showed that Flock employees used integrated private-camera feeds during product demonstrations. The feeds included views inside or around sensitive locations such as a children’s gymnastics room, a pool, fitness areas, a school and a playground.
These were not Falcon roadside plate-reader images. They were privately owned camera feeds connected to the city’s Flock platform.
Flock said access was authorized under its agreement with the demonstration partner and denied that employees were “spying on children.” The company’s CEO apologized for poor judgment, and Flock said it stopped using sensitive Dunwoody locations and real customer feeds for those demonstrations.
That response matters. So does the event itself. A resident deciding whether to approve an integrated platform should know that “the customer owns the data” does not answer every question about vendor access, authorized support, demonstrations or connected private feeds.
404 Media: Dunwoody learned sensitive camera feeds were used in Flock demonstrations ↗
Flock’s current FAQ on data access ↗
A Flock demonstration environment searched real customer cameras.
In September 2026, records-based reporting described a company-controlled demonstration environment known as “Flock City PD.” Audit logs showed its accounts running FreeForm queries against real cameras in Dunwoody, Georgia and Bryan, Texas while testing or demonstrating searches involving crowds, masks, signs and political or religious stickers.
Flock said the activity involved moderation testing and demonstrations rather than customer criminal investigations. The company said some sensitive prompts generated warnings or were blocked, and it later isolated the environment so it no longer searched real customer cameras.
The correct conclusion is not that every sensitive query succeeded or that local police initiated them. It is that real customer imagery was used in company testing or demonstrations, and natural-language search was tested against characteristics that can reveal political, religious or associational activity.
404 Media: “Flock City PD” searched real cameras during testing and demonstrations ↗
Condor feeds and archives were exposed without a login.
On December 22, 2025, 404 Media reported finding at least 60 internet-exposed Condor camera interfaces that did not require authentication. Reporters could view live feeds, retrieve roughly 30 days of recorded material and, in some instances, reach settings, logs or other administrative functions. The interfaces were publicly discoverable through Shodan, an internet-device search service.
The views included children on a playground, a woman on a path and people in parking or roadside areas. Reporters observed cameras automatically zooming toward or following people. One reporter visited a camera and watched himself appear live while colleagues viewed remotely. Benn Jordan also demonstrated how exposed imagery could be combined with facial-search and other open-source information to identify people and learn personal details.
Flock characterized the incident as a limited deployment misconfiguration affecting a small number of cameras and said it remediated the problem.
This is not evidence that every Condor camera was exposed or that Flock’s central cloud database was breached. No reported abduction resulted. It is a documented security incident involving Flock video-camera deployments and public access that could reveal who was present or alone, routes, routines and identifying details—creating genuine stalking and child-safety risks.
404 Media: exposed Condor feeds and archives ↗
WPTV: Flock confirms and responds to the exposed feeds ↗
Researchers extracted data from a physically removed camera.
WIRED reported in September 2026 that people who physically removed and examined a Flock camera extracted device material and imagery from it. The reported files included roughly three weeks of images and clips and software detections involving vehicles and some people. The report did not establish active facial recognition.
Flock said unauthorized removal is illegal and criticized the researchers for not using its disclosure process. We agree on the essential conduct point: do not touch, damage, remove, access or interfere with surveillance equipment.
The lawful policy lesson is narrower. Physical security and device-level storage are part of the threat model, and broad claims that nobody could obtain imagery through physical device access should be tested against documented technical evidence.
WIRED: data recovered from a physically removed Flock camera ↗
“Flock has never been hacked” needs a precise answer.
Flock says it has not experienced a verified compromise of its central cloud platform involving customer-data exfiltration. We have not established evidence that contradicts that narrow claim.
Flock: “Has Flock been hacked?” ↗
That is not the same as saying there has never been a meaningful security failure involving a Flock product. The exposed Condor deployments and reported device extraction are concrete security events. A careful answer separates:
- compromise of the central cloud platform;
- exposure caused by a camera or deployment configuration;
- compromise or extraction after physical device access;
- a vulnerability demonstrated in controlled research; and
- misuse of valid user credentials.
Calling every one of those a “cloud hack” would be inaccurate. Calling all of them harmless because the central cloud was not breached would also be inaccurate.
Independent researcher Jon “GainSec” Gaines published a laboratory-based assessment covering Flock hardware, embedded software and apps. The project reports 51 findings, including 22 assigned CVEs and eight awaiting assignment at publication. The researcher says the work 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. The company said the findings generally required physical device access and specialized hardware knowledge, said it had no indication of customer exploitation and described updates or mitigations in progress. Those responses should be part of the record too. A vulnerability is evidence of a weakness, not proof that somebody used it against a customer.
Primary research repository and disclosure notes ↗
Archived research paper on Zenodo ↗
Flock’s response to the compiled device research ↗
Accessible secondary overview from SimeonOnSecurity ↗
Searches have involved protests and political activity.
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.
The records do not prove that every query was unlawful or that every agency used the same method. They do show that protest-related searching is not merely a hypothetical future use.
The Record: EFF analysis of protest-related Flock searches ↗
This matters even when a plate is visible on a public road. A searchable list of vehicles near a demonstration can reveal association and participation at a scale that casual observation cannot.
Searches have crossed jurisdictions for highly sensitive investigations.
Associated Press reporting documented a Texas search for a woman who had self-administered an abortion that reached a network of roughly 83,000 cameras. An Illinois state audit later found Flock-related access that officials said violated state restrictions involving abortion, immigration and gender-affirming care. The state ordered changes.
Associated Press: abortion investigation reached a nationwide camera network ↗
Illinois Secretary of State: official audit findings ↗
Flock and local agencies describe sharing as controlled by customer choices and permissions. The Illinois and Ventura records show why residents should verify actual settings, indirect request paths and audit history instead of relying only on the intended policy.
Authorized access has been used for alleged personal surveillance.
Public records and reporting have documented officers or investigators accused of using Flock or related law-enforcement databases to search vehicles connected to former romantic partners. Some cases remain allegations or charges; they should not be described as final convictions unless the record supports that.
The point is not that every officer behaves this way. It is that a valid credential, required search reason and audit log do not make an improper search impossible.
Read the sourced misuse and false-match cases →
A Metro Detroit camera was removed from a sensitive location.
Madison Heights required a Flock camera at the entrance to JARS Cannabis as part of an agreement allowing the business to operate. The business said it did not control the camera or have access to its data. After customers raised privacy concerns and the business protested the placement, the city and JARS reached a revised agreement and the camera aimed at the parking lot was removed in August 2026.
Another nearby street-facing Flock camera remained. This was removal of one camera, not abolition of Madison Heights’ wider system.
The example matters because location changes meaning. A camera aimed at the entrance to a cannabis dispensary can reveal visits to a legally operating but sensitive business. The outcome also shows that public pressure can change a local surveillance decision.
Local 4: Madison Heights and JARS reached an agreement to remove the camera ↗
FOX 2 Detroit: community debate and planned council vote ↗
What about the camera a critic says appeared near his driveway?
Attorney and video creator Jake the Lawyer published a first-person account describing a Flock camera installed near or aimed toward his driveway after he began researching the company.
The timing and placement are part of his account. They do not by themselves establish who selected the location or prove that the camera was installed in retaliation for his work. We have not found independent evidence proving retaliation.
That is the line between a relevant personal account and a verified motive. We can discuss the placement concern without turning chronology into proof.
Watch the video and read our source note →
Flock actively helps agencies answer public objections.
Flock does not leave every police department or city official to explain the product alone. Its own webinar for law-enforcement communications advertises a social-media toolkit, a chief-to-council guide and an “objection-handling playbook” intended to help agencies reframe common concerns and maintain public support.
Flock webinar: How to Speak to City Councils and Communities About Flock ↗
That does not prove a secret cover-up, and it does not make every statement from an agency false. It establishes that the vendor has a direct role in shaping how customers present the system and respond to criticism.
A vendor presentation is advocacy from a company that wants the contract. Residents should compare it with the signed agreement, product entitlements, sharing settings, audit logs, incident notices, independent reporting and the experience of communities that ended their systems.
“The data disappears after seven days.”
Seven days is shorter than 30 days, and shorter routine retention reduces the amount of historical data immediately available. That is a real improvement while a system exists.
It is not the same as saying every copy disappears after seven days.
Flock’s current evidence policy says routine raw LPR evidence is generally retained for seven days unless a different period applies. It also describes customer downloads and exports, preservation for an active matter and approved extended retention. Once information enters a screenshot, report, intelligence bulletin, court filing, local evidence system or another agency’s file, the original vendor deletion period may not control that copy.
Read the full sharing and retention guide →
Common objections, answered without exaggeration
“I drive in public. I have no expectation of privacy.”
Being visible once is not the same as being automatically recorded, indexed and searched across time and place. Privacy includes practical obscurity: ordinary people cannot usually remember and combine every public movement. A networked database can.
“If you have nothing to hide, you have nothing to fear.”
Privacy is not an admission of wrongdoing. People reasonably protect medical visits, worship, political activity, intimate relationships, legal consultations and their children’s routines. Laws, administrations, policies and personal circumstances change. A database created today can be used under tomorrow’s priorities.
“No facial recognition means it is safe.”
Facial recognition is one risk, not the definition of surveillance. Plates, vehicle features, clothing, time, place, companions, repeated routes and visits can reveal identity and association without a biometric face match.
“A few bad users do not justify throwing out the whole system.”
The objection is not only that a few users may break rules. The system’s ordinary purpose is to collect and make searchable observations about large numbers of people who are not suspected of a crime. Misuse, mistakes, access failures and security incidents are additional reasons to reject that underlying architecture.
“Audit logs prevent abuse.”
Audit logs can deter, expose and help discipline misuse. They usually operate after access exists. A log can show that someone searched for an ex-partner; it cannot make the search unhappen or guarantee that an exported result was not copied.
“Sharing is optional, so our data is local.”
Optional sharing makes the local setting important; it does not tell us what the setting is. Ask for the partner list, network permissions, configuration history, indirect request procedures and search audits.
“The cameras solve crimes.”
They sometimes provide useful investigative leads. The policy question is whether those benefits justify collecting searchable location information about ordinary traffic first. Targeted investigation, warrants, relevant business video, witnesses, forensics, tips and other tools remain available without constructing a routine regional movement database.
“Our officers would never abuse it.”
Public policy should not depend on perfect users forever. Good people make mistakes; accounts can be misused; permissions can be configured incorrectly; leadership changes. The issue is the durable power created for whoever holds access next.
“We can regulate it instead of removing it.”
Stronger rules, shorter retention and better audits can reduce harm while a system exists. They do not answer the threshold question: should government routinely collect the location observations in the first place? Our conclusion is no; the safer end state is removal and nonrenewal.
“Phones, toll systems and doorbells already collect data.”
One invasive system does not justify another. Different systems have different operators, consent models, purposes, legal rules and access paths. The spread of commercial and government surveillance is a reason to limit additional collection, not surrender to it.
What we will not claim
Credibility is part of the campaign.
We will not claim that:
- every Falcon plate reader searches for people;
- Flock’s current products have been shown to perform facial recognition;
- every municipality purchased or enabled every advertised feature;
- a camera appearing near a critic proves retaliation;
- the canceled Ring integration transferred customer video to Flock;
- every protest-related query was unlawful; or
- a confirmed central-cloud breach has been established when the evidence here does not establish one.
Correcting those overstatements does not weaken the case for removal. It makes the documented case harder to dismiss.
Read the accurate history of Flock’s canceled Ring integration →
The policy conclusion
The public is often asked to evaluate these systems under ideal conditions: trained users, correct configurations, short retention, narrow sharing, perfect cybersecurity, unchanging laws and a vendor that always exercises good judgment.
The real record includes mistakes, misuse, sensitive searches, product expansion, vendor access, changing permissions, exposed feeds and copied information.
You do not have to believe every worst-case scenario will happen in your city. You only have to decide whether the benefit of this collect-first infrastructure justifies creating the power and carrying the risk.
Our answer is no.
The safest movement database is the one the government does not build.
One useful action
Ask your municipality for:
- the current contract, order forms, product list and renewal deadline;
- every camera, sensor, software module and private-feed integration;
- current retention and preservation settings;
- sharing partners, national or statewide network settings and indirect-request procedures;
- user roles, audit logs and documented misuse or security incidents;
- vendor communications, presentation materials and sales demonstrations; and
- records showing what happens to local data and accounts if the contract ends.
Then ask the officials with lawful authority to place nonrenewal or termination on a public agenda before the deadline.