A license plate is visible in public.
So is a car.
So is a person walking down the sidewalk.
The difficult question is not whether somebody could see you in public.
It is what happens when technology can record, store, search, connect and analyze observations at a scale people could not practically reproduce.
Privacy is not the same as secrecy.
We close restroom doors, draw curtains, use passwords and choose who receives personal information. Most people do not publish a minute-by-minute log of everywhere they go.
None of that means they are doing something wrong.
Privacy is about reasonable limits on who can collect, search and combine information about us. Being visible at one moment in public is different from having many observations automatically indexed so they can be searched later.
“Cameras are everywhere” does not make every system equivalent.
| Technology | Typical privacy issue | Why it is not automatically equivalent to ALPR |
|---|---|---|
| Ordinary CCTV or webcam | Records video of a place or scene. | It may require manual review and may not automatically index a vehicle into a location-history search system. |
| Phone location data | Can reveal extensive location history and raises serious privacy concerns. | The data holder, legal process, controls and collection mechanism differ. Its invasiveness does not justify more surveillance. |
| Data brokers | Aggregate and sell or share personal information. | A separate serious privacy problem is not a reason government should create another searchable movement dataset. |
| Smart TV, assistant or doorbell camera | May collect viewing, audio, device or household information depending on the product and settings. | The purpose, control and legal context differ. The common lesson is that collection deserves scrutiny. |
| ALPR network | Creates searchable vehicle observations with time, location and other visible attributes. | It is purpose-built for retrospective vehicle search and may support interagency or wider network access. |
Start with the five-minute ALPR introduction →
It is not really about the license plate.
A plate number is an identifier attached to a vehicle. A vehicle is usually connected to a person.
Over time, vehicle observations can point toward:
- where someone works;
- where they spend nights;
- which medical offices they visit;
- which religious institutions they attend;
- political or community meetings;
- friends and family;
- recurring routines;
- and other places that may reveal far more than a license plate alone.
The plate can be the entry point. The privacy issue is the picture that can emerge around it.
One trip tells very little. A pattern can tell a lot.
One record saying your vehicle passed one intersection at 4:17 PM may be boring.
Now imagine hundreds of records across many locations and days.
That is sometimes described as a mosaic problem.
A single tile reveals little. Enough tiles reveal the picture.
Public does not automatically mean harmless when technology can remember so much for you.
Seeing is not the same as tracking.
A police officer can stand on a road and see cars pass.
What that officer cannot practically do is stand at hundreds of roads simultaneously, remember every vehicle for weeks, index all those observations and answer a historical query in seconds.
Automation changes the scale.
That difference is central to the privacy question.
Being watched can change lawful behavior.
Researchers often use the term Hawthorne effect for behavior changes that occur when people know they are being observed. A systematic review found evidence of an effect in some studies but also warned that the term, methods and measured effects vary. A newer review focused on video recording found that some participants reported behavior changes, while stressing substantial uncertainty and study limitations.
In a civic setting, chilling effect is often the more useful phrase. A person may think twice about attending a lawful protest, visiting a clinic, entering a place of worship, meeting a source or having an ordinary private conversation when they believe an identifiable record may be watched, searched or reviewed later.
That does not mean every camera changes every person’s behavior, or that every observed change is harmful. It means pervasive observation can impose a real social cost even when no one is suspected of a crime. Freedom is not only the absence of punishment; it includes being able to take lawful actions without constantly calculating how a permanent audience might interpret them.
Systematic review of the Hawthorne effect →
Systematic review of behavior change around video recording →
The courts have recognized the difference in other location technologies.
The United States Supreme Court has not declared Flock or ALPR networks categorically unconstitutional.
That claim would be inaccurate.
But important cases have recognized that aggregated digital location tracking can become constitutionally different from isolated observation.
United States v. Jones involved prolonged GPS tracking. The majority decided the case through the physical installation of the tracker, while concurring Justices also warned about the revealing nature of long-term automated monitoring.
Carpenter v. United States held that obtaining the historical cell-site location information at issue was a Fourth Amendment search. The Court recognized how revealing a detailed location history can become.
Chatrie v. United States, decided by the Supreme Court in 2026, held that government acquisition of the cell-phone Location History information at issue through a geofence process was a Fourth Amendment search, reinforcing that digital location information does not become insignificant merely because a technology company possesses it.
Leaders of a Beautiful Struggle v. Baltimore Police Department held that accessing Baltimore’s accumulated aerial surveillance data was a search because the system enabled police to deduce information from the whole of individuals’ movements.
ALPR-specific cases remain fact dependent and mixed. That is why our legal page shows the cases that cut both ways.
Read the Fourth Amendment & Location Tracking page →
The technology keeps becoming more powerful.
Modern ALPR platforms can go beyond exact plate searches.
Flock publicly markets searches using vehicle characteristics, natural-language descriptions, shared camera networks and newer tools intended to connect sightings and identify patterns between vehicles.
Separate video products can search people using visible, non-biometric characteristics.
That does not prove every feature is enabled in every municipality.
We keep four questions separate:
- Vendor capability - what the company says the product can do.
- Local configuration - what a particular agency has purchased or enabled.
- Written policy - what the agency says users may do.
- Actual use - what records show users actually did.
Those are not interchangeable.
But vendor capability still matters. A feature that is not documented as enabled today may become available through a subscription, software change, integration or permission change without replacing the roadside hardware. Unless the public can inspect current entitlements, configuration changes and audits, residents may not know when that happens.
Our rule is: do not present technical possibility as proven local use—and do not pretend an unverified present configuration defines the system’s future risk.
See how much these systems can actually see →
Sharing can make a local camera much less local.
A municipality may own only a small number of cameras.
The more important question can be how much outside data its personnel can search and who can search the municipality’s own data.
Sharing arrangements, permissions and integrations matter.
A local procurement decision can become part of a much larger surveillance network.
Access can be abused.
We do not argue that every officer abuses surveillance access.
We do not need to.
Powerful databases create risks because people make mistakes, policies are violated, permissions are misconfigured and insiders sometimes misuse legitimate access.
There are documented ALPR and law-enforcement database cases involving searches connected to former romantic partners, sensitive out-of-state investigations and other misuse.
Audit logs can help discover abuse after the fact.
They do not make the database disappear.
See documented misuse, failures & harm →
Technology can also be wrong.
Automated systems can misread plates, rely on stale information or produce matches that require careful verification.
A database result can look authoritative because it has a timestamp, location and image.
That does not turn it into infallible evidence.
When an automated result is wrong, an innocent person can experience the consequences.
Seven days can still reveal a pattern.
Flock’s current public website says it reduced its recommended/default routine retention from 30 days to seven days. Its current privacy materials say most communities use seven days unless local law says otherwise.
Some current Flock legal material still contains older 30-day default language. That is another reason we do not assume a universal retention period for any Metro Detroit agency. We verify the local contract, policy and configuration.
Seven days is a meaningful reduction from 30.
It is still enough time for repeated trips and routines to form a pattern.
Think about your own last week.
How many trips did you make?
How many routines repeated?
Which places would reveal something meaningful about your life?
The question is not whether seven days is better than 30.
The question is what can be learned from routine observations during that period, and whether those observations should be collected first.
Flock Safety: data privacy and retention →
“I’m not doing anything wrong.”
Privacy is not a reward for wrongdoing.
You probably close your curtains without committing a crime.
You probably do not publish every medical appointment, relationship, conversation or routine online.
And you probably have at least one completely lawful trip you would rather not turn into a family group-chat discussion.
The question is not whether you have something criminal to hide.
It is whether government should routinely build a searchable movement history on people who are not suspected of anything.
But these cameras can help solve crimes.
Sometimes they do.
That does not end the discussion.
The right question is whether that benefit justifies the collection model.
Police have many other investigative tools: witnesses, physical evidence, warrants, security footage, forensics, public tips, lawful digital evidence, open-source research and ordinary detective work.
We support effective investigation.
We oppose collecting routine movement records first and searching them later when a reason appears.
Read Public Safety Without Mass Tracking →
Privacy and public safety are not a binary choice.
We can want stolen cars recovered and still oppose mass tracking.
We can respect police officers and still question a surveillance contract.
We can support modern investigative technology and still decide a particular technology goes too far.
Democratic government is supposed to make those tradeoffs openly.
Regulation is not our endpoint.
Shorter retention, stronger auditing, tighter access controls and warrant requirements may reduce harm while a system exists.
If a system exists, stronger safeguards are better than weaker ones.
But the records are still collected.
The database still exists.
Our campaign goal is complete removal and nonrenewal of covered mass vehicle-surveillance systems.
The useful part is that these systems are not inevitable. They are purchased, funded, renewed and governed through local decisions.
A renewal date is a decision point, not an inevitability.
You do not have to solve the whole regional network. Start with the system, contract and next decision in your own community.
We show our work.
We have a policy position. We do not pretend to be neutral about whether these systems should remain.
But our factual claims still need to be accurate.
If a system helped solve a crime, we can acknowledge it.
If a court rejected an ALPR challenge, we can show that too.
If we cannot prove a local feature is enabled, we say so.
If something is unknown, we label it unknown.
Our argument should not require bad information.
The documented reality is concerning enough.