Second Report of the Axon AI

& Policing Technology Ethics Board:

Automated
License Plate
Readers
OCTOBER 2019
Table of Contents
I EXECUTIVE SUMMARY 4

II ALPRs TODAY & AXON’S ENTRY INTO THE ALPR MARKET 10

A. How ALPRs Work 12
B. ALPR Use Today 13
C. Axon’s Entry into the ALPR Market 14

III PRELIMINARY ASSESSMENT AND GUIDANCE

REGARDING ALPR BENEFITS & COSTS 16
A. Preliminary Assessment of Potential ALPR Benefits 18
1. Improved Real-Time Location of Hot-List License Plates 18
2. Improved Historical Investigations 19
3. General Crime Deterrence 20
4. Improved Traffic Safety 20
5. Increased Revenues 20
6. Reducing Arbitrariness and Bias 20
7. Officer Safety While Driving 21
8. Predictive Analysis 21
B. Preliminary Assessment of Potential ALPR Costs 22
1. Increased Enforcement Generally and Its Likely Disparate 22
Racial, Ethnic, and Socioeconomic Impacts
2. Increased Fines-and-Fees Enforcement and Its Disparate Impacts 23
3. Inappropriate Stops 23
4. The Privacy Implications of Growing ALPR Databases 24
5. Additional Potential Costs Depending on Use 25
6. Lack of Information and Transparency Around ALPR Use 26

2
IV RECOMMENDATIONS 28

V FUTURE STUDY & STAKEHOLDER ENGAGEMENT 38

A. Continued Research 39
B. Stakeholder Engagement 41
C. Model Statute & Model ALPR Policy 41

VI CONCLUSION 42

VII ATTACHMENTS 44
A. 10 Questions to Ask Your Law Enforcement Agency About ALPRs 44
B. Product Evaluation Framework 47
1. Guidance on Assessing Potential Benefits 47
2. Guidance on Assessing Potential Costs 48
C. References 49

3
I
Executive
Summary

4
In 2018, Axon Enterprise, Inc. established the AI and Policing Technology
Ethics Board. Its purpose is to help guide and advise the company on
ethical issues relating to its development and deployment of new artificial
intelligence (AI)-powered policing technologies that are under consideration
or development, not to formally approve particular products. The Board’s
first publication, First Report of the Axon AI & Policing Technology
Ethics Board (“First Report”), was released in June 2019 and provided
information about how the Board operates, as well as recommendations
on the use of face recognition technology in Axon products.
This is the Board’s second report and concerns
automated license plate readers (“ALPRs”).
ALPRs represent one of the most widely is insufficient regulatory oversight. Indeed,
used surveillance systems in existence. while there is considerable concern about
ALPRs are computer-controlled camera face recognition, the subject of this Board’s
systems that read passing license plates. First Report, widespread face recognition
As used by law enforcement, ALPRs alert use by law enforcement remains, for the
officers when they encounter a license plate most part, hypothetical. ALPRs, on the other
of interest—such as the license plate of a hand, are ingrained into law enforcement
vehicle that has been reported stolen. ALPR operations across the country. They are
systems typically store data on all of the here today, and they are very real.
license plates they read, creating extremely Nonetheless, we lack information about key
large databases of vehicle information over aspects of law enforcement’s ALPR use. We
time (also known as “historical data”). do not know how many ALPRs there are,
The use of ALPRs has become common where they are deployed, how effective they
across the private and public sectors—in are at helping to catch serious criminals,
everything from public toll collection to or how frequently they are used simply to
private surveillance and security—but there enforce low-level offenses. Retention rates

5
for ALPR data vary widely, yet we largely are
ignorant of how that historical data is used
and accessed. Within this report we cite
evidence indicating that the impact of ALPRs
falls disproportionately on lower-income
individuals and people of color, but again, this
impact has not been quantified or assessed.
From what the Board knows, we can say
that the potential benefits of ALPRs appear
The use of ALPRs
real, yet the extent to which these potential has become common
benefits are realized is unknown. ALPRs can across the private
be used to locate stolen vehicles, identify
vehicles of interest with regard to serious
and public sectors—
crimes, and locate the subjects of Amber or in everything from
Silver alerts more efficiently and effectively. public toll collection
Police can use stored ALPR data to conduct
investigations into a wide variety of crimes.
to private surveillance
Yet, it also is clear that ALPRs impose very
and security—but
real costs. ALPR-aided enforcement may there is insufficient
exacerbate racial and socio-economic regulatory oversight.
disparities in the criminal system. This is
particularly the case when, as we document
below, ALPRs are used to enforce low-
level traffic offenses or generate revenue
design. Axon also suggested a number
in the form of fines and fees. ALPRs can
of possible guardrails on its ALPRs.
generate false positives that may lead to
intrusive and potentially dangerous traffic Axon’s entry into ALPRs comes at a pivotal
stops. ALPR systems store data, often for moment. ALPR prices already have dropped
long periods of time, creating historical sharply and will continue to decline, making
databases of our activities that for the most widespread adoption seemingly inescapable
part are unregulated, creating serious privacy whether Axon enters the market or not. At
risks. ALPRs and the data they store can the same time, the market is not consolidated
and have been used in ways that chill First and each player is following its own rules.
Amendment liberties and threaten Fourth Without regulatory intervention, there
Amendment rights. The full extent of the is a risk that competition will encourage
costs of ALPRs are unknown as well. a race-to-the-bottom of more pervasive
and more powerful surveillance.
Axon has informed us that it intends to enter
the ALPR market in the near future, providing The Board hopes to help guard against a race-
low-cost AI-driven ALPR technology as a to-the-bottom in the use of ALPRs. Although
part of its dashboard cameras. Consistent some jurisdictions have laws governing ALPR
with a recommendation of our First Report, use, most do not. Similarly, many agencies
Axon notified us of its plans early, giving us have policies addressing key issues around
an opportunity to influence the product’s ALPRs, but many do not, and of those that

6
do, the policies are not always publicly will enter. The extent to which Axon follows
available. In this report, we call for immediate our recommendations in this report will bear
regulation in all jurisdictions in which ALPRs on our assessment of whether its product
are in use, including self-regulation by and its usage are appropriate and ethical.
the industry and law enforcement, and— As we said in our First Report, communities—
critically—government regulation as well. not law enforcement agencies—are the
(We similarly believe the private usage of ultimate customers for technology purchased
ALPRs should be regulated, but view that by government. Not only should communities
as beyond the scope of our review.) have a say in what surveillance technology
In addition to calling for regulation of ALPRs, gets purchased and how it is used, but law
this report includes specific recommendations, enforcement agencies must adopt policies
both to inform law enforcement’s use of ALPRs and practices that ensure their use of ALPRs is
around the country and to offer concrete equitable, well-informed, and transparent to
ethical guidance to Axon as it pursues its the communities they serve. Vendors should
ALPR. In making these recommendations, design products to help ensure that—to the
we do not necessarily endorse ALPR use extent this technology is deployed—it is used
generally, nor Axon’s decision to enter this transparently, equitably, and effectively. To
market. Because Axon’s ALPR remains under this end, we have included, in Attachment A,
development and a release date has not yet a document the Board prepared that offers
been determined, we will provide specific suggested questions community members
views to Axon about its product when we learn can ask about ALPR use in their jurisdiction.
more about it and the state of the market it

7
As described in much more detail in the pages that follow, we
have provided specific recommendations to help guide Axon,
law enforcement, and the ALPR industry generally. These
recommendations and their rationale are spelled out in detail
in Part IV, and we summarize here:

01 Law enforcement agencies should not acquire or use ALPRs without going
through an open, transparent, democratic process, with adequate opportunity
for genuinely representative public analysis, input, and objection. To the extent
jurisdictions permit ALPR use, they should adopt regulations that govern such use.
(This is what we said about face recognition, and it is true as well for ALPRs.)

02 Agencies should not deploy ALPRs without a clear use policy. That policy should be
made public and should, at a minimum, address the concerns raised in this report.

03 Vendors, including Axon, should design ALPRs to facilitate transparency about their
use, including by incorporating easy ways for agencies to share aggregate and de-
identified data. Each agency then should share this data with the community it serves.

04 Vendors, including Axon, should design their ALPRs so that agencies can adjust
the list of vehicles to which an ALPR will alert law enforcement officers, so that
the list includes only those offenses or reasons most of concern to that agency
and its community. Although communities must decide the contours of their
own alert lists (also known as “hot lists”), as a general matter, we believe that
these lists should not be used to enforce civil infractions, offenses enforceable by
citations, or outstanding warrants arising from a failure to pay fines and fees.

05 Vendors, including Axon, must provide the option to turn off immigration-
related alerts from the National Crime Information Center so that jurisdictions
that choose not to participate in federal immigration enforcement can do so.

06 ALPRs must be designed and operated in ways that ensure alert

lists are checked routinely for errors and kept up to date.

8
07 An ALPR alert, on its own, should not constitute sufficient grounds to stop a vehicle.
Officers must make visual confirmation independently that the license plate matches
the hot-listed plate. If the offense at issue is associated with the registered owner
of the vehicle (as opposed to the vehicle itself), the officer also should ascertain
whether the driver is consistent with the description of the registered owner.

08 Axon should work with partner agencies to determine the shortest possible
retention period for ALPR data that will serve law enforcement needs
sufficiently, and set that period as the default retention setting on its ALPRs.

09 ALPR design should create audit trails both of real-time ALPR

alerts and agency accessing of historical data. Law and agency
policy should require regular auditing of ALPR usage.

10 Stored ALPR data must be encrypted and secured against outside access and breach.

11 ALPR vendors should not retain the right to access or share ALPR data, and law
enforcement’s ALPR data never should be shared for use by for-profit third parties.

12 ALPRs should be designed such that if agencies share data with other
law enforcement agencies, they do so transparently and in a way that
is governed by formal and lawful data-sharing agreements.

13 Vendors, including Axon, never should profit from fines and

fees obtained through law enforcement use of ALPRs.

14 Vendors, including Axon, should provide adequate training materials

for agencies and officers using their ALPRs, including about
default settings and why they are set the way they are.

15 It is imperative that data-gathering and impartial study be conducted of

ALPR usage, so that communities and the country are aware of how ALPRs
are being used, of what is required to make that usage effective, of any
harms arising from ALPR usage, including whether ALPRs are exacerbating
disparities, and of ways to eliminate or mitigate those harms.

9
II
ALPRs Today &
Axon’s Entry into
the ALPR Market

10
 Should Axon enter this market, we call
on it to be a leader on this front—to
design its product with controls to assure
transparency, equity, and other ethical
principles are in place. We hope that if one
prominent vendor does this, communities
will take notice and insist their agencies
use products with similar controls.
Despite the fact that ALPR use has exploded The way to stop a race-to-the-bottom is
internationally and nationally over the last two with regulation—including industry self-
decades, in many places in the United States, regulation and government regulatory
ALPR usage is unregulated or under-regulated. intervention.1 Should Axon enter this market,
This relative lack of regulation, combined with we call on it to be a leader on this front—to
rapid growth, has created an industry with design its product with controls to assure
little public accountability. Literally billions of transparency, equity, and other ethical
license plate “reads” of ordinary individuals principles are in place. We hope that if one
sit in ungoverned databases, leading to prominent vendor does this, communities
enforcement stops and allowing for historical will take notice and insist their agencies use
tracking. It is difficult for law enforcement to products with similar controls. But we also
resist the allure of these tools, and vendors realize Axon is not alone in the ALPR market
are incentivized to outdo one another with and cannot alone dictate its terms. Therefore,
regard to what capacities are built into ALPRs. this report includes an urgent and emphatic
call for government regulation of ALPRs.

11
How ALPRs Work
ALPRs are computer-controlled camera systems that can be mounted almost anywhere.
They can be stationary or mobile. They can be attached to light poles, police cars, private
tow-trucks, or even mobile phones. Any ALPR system has four essential components:

01 | Capture:
ALPR cameras are designed to record license plates by automatically
capturing video or images of passing vehicles. Some ALPRs use
infrared imaging, while others (often newer) operate with normal
visual-spectrum imaging from a high-definition video camera.

02 | Compute:
An algorithm detects license plates in the captured image, and “reads”
the characters of the license plate. Some ALPRs also are able to detect
information about the vehicle, such as color or type. Because the camera
captures an image of the passing vehicle (and not just the license
plate), in the future, algorithms may be able to identify yet additional
information, including occupants of the vehicle through face recognition.

03 | Catalog:
Most ALPR systems then “catalog,” or store, the license plate number,
time, and GPS-coordinates of each scan. This allows the tracking
of where vehicles have been over time through stored data.

04 | Compare:
Most, though not all, ALPRs then compare license plates they come across
against a database of sought-after plates—commonly referred to as an
“alert list” or “hot list.” If the ALPR detects a match with the hot list, the
ALPR user—usually an officer in a patrol car—is notified in real time.

12
There are both national and local hot lists. Over time, ALPR use has exploded along
The national hot list, from the National Crime several dimensions.
Information Center (NCIC) maintained by
According to one estimate, between 2007 and
the Federal Bureau of Investigation, contains
2012, the number of U.S. law enforcement
license plates associated with vehicles of
agencies reporting ALPR use jumped from
interest in a wide variety of concerns and
19% to 71% with 85% planning to buy or
offenses. States and municipalities can opt in
expand their systems.4 ALPR use is not
to receive NCIC alerts and can generate their
limited to government and law enforcement.
own local hot lists as well. State and local hot
Certain ALPR vendors also sell their systems
lists may include license plates associated
and databases to private users, including
with lower-level enforcement concerns, from
repossession companies, banks, and insurance
individuals with local bench warrants to plates
companies.5 Others vendors focus on
associated with outstanding parking tickets.
neighborhood and home security, bundling
cameras with installation, maintenance,

ALPR Use Today

and data-retention services.6 ALPRs also are
widely used for toll collection. Overall, the
estimated value of the global ALPR market
ALPRs have been used by law enforcement in 2018 was $1.81 billion, with the projected
agencies around the globe for over 20 years. value to increase to $3.43 billion by 2025.7
The first large-scale implementation of ALPRs
This expansion has been fueled both
was in 1997 in the United Kingdom. ALPRs by government and by private industry.
were integrated into London’s “ring of steel,” Federal and state funding, including federal
a network of CCTV cameras mounted around grants, provided tens of millions of dollars
the perimeter of the central district of London, toward ALPR purchases and allowed many
tracking every vehicle coming in or out. The agencies to purchase their first ALPRs.8
following year the U.S. Border Patrol began to Since then, private vendors have pushed
use ALPRs to track vehicle movement at the ALPR costs lower and lower. Technological
U.S. border.2 advancements have allowed off-the-shelf
ALPRs first were used in the 1990s by local U.S. digital cameras to be converted into ALPR
law enforcement to combat automobile theft. scanners.9 One company, which offers its
The idea was that if a stolen vehicle (or, more software for free to run on ordinary security
accurately, a vehicle with a license plate on a cameras (customers can optionally purchase
hot list of stolen vehicles) passed an officer, the cloud storage and analytic services), recently
officer would be alerted and could investigate.3

13
reported a nearly 3,000% increase in agency terminates.13 Law enforcement agencies
downloads and use from 2017 to 2019.10 have contracts with vendors that grant access
to the private databases and facilitate sharing
ALPRs also have expanded from their
license plate reads with other agencies.14
original purposes. Initially, ALPRs primarily
were used to locate stolen vehicles. Today, Today, ALPR databases store information
ALPRs can connect to external databases about the locations of the vehicles of millions
to learn more about the license plate or the of people who are innocent of any offense.
vehicle’s registered owner, alert an officer The current market leader in historical ALPR
to a vehicle’s unpaid traffic tickets or expired data boasts that it has a database of 9 billion
insurance, or flag the registered owner’s nationwide plate scans from private companies,
open warrants, unpaid property taxes, sex- along with the ability to access 1.5 billion plate
offender status, or supposed gang affiliation.11 scans from law enforcement agencies.15
Importantly, today ALPR scans typically

Axon’s Entry into

are stored and retained, including time
and location information, creating vast
databases allowing for investigations
into where particular cars have been over the ALPR Market
time.12 We refer to the information in
these databases as “historical data.” For decades, the cost of mobile ALPRs served
as a natural limit on adoption. Most agencies
were unable to afford to equip more than a
Today, ALPRs can connect to few cars at a time.16 The most common mobile
external databases to learn ALPRs consist of multiple cameras mounted
more about the license plate externally on a vehicle. Given the expenses of
installation, the need to make these cameras
or the vehicle’s registered
weather resilient and tamper-proof, and the
owner, alert an officer to a underlying hardware and technology costs,
vehicle’s unpaid traffic tickets early models retailed at nearly $25,000
or expired insurance, or flag per vehicle. But, as noted above, new
the registered owner’s open technologies are causing the price of ALPRs to
warrants, unpaid property drop precipitously. Vendors are now making
taxes, sex-offender status, or ALPR systems, powered by AI-technology,
supposed gang affiliation. that can run on almost any camera.
It is in this context that Axon informed
These databases of historical data are amassed the Board that it plans to enter the ALPR
through a variety of methods. Private vehicles market. Axon intends to make ALPR
(often repossession trucks) travel the streets technology available on a future version
equipped with ALPRs. Police cars gather of the dashboard cameras that exist
license plate reads as they patrol, storing the in many police patrol cars already.
reads in agency systems. Vendors enter into Axon already produces patrol car dashboard
contracts with law enforcement agencies cameras (“dash cams”) under the Axon
that allow access to all agency plate scans, Fleet brand name.17 Axon expects its next
and retention even after the contract with the generation Fleet camera to include ALPR

14
capabilities. Axon’s ALPR will use AI image across an agency’s entire fleet of cars.
processing and machine learning to extract
Below, we have included the information we
license-plate data from video, which will be
have so far about Axon’s proposed ALPR. We
fed into a terminal (usually a laptop) that
expect to be provided with additional details in
officers have in their cars. By integrating ALPRs
time. In addition to ironing out engineering and
into dash cams (as opposed to externally
design issues, Axon faces a range of choices
mounted cameras), Axon expects that its
that will impact how its ALPR can be used by
innovations will enable agencies to deploy
law enforcement agencies, including data
ALPRs more broadly than before, potentially
ownership, data access, privacy, and security.

What we know about Axon’s ALPR:

Integrated with in-car cameras,
either dash cams or rear cams;

Equipped with 60-degree field of view allowing

for scanning three lanes of traffic;

Accuracy of plate reading algorithm approaches

100% under ideal conditions;

Accuracy decreases depending on lighting,

distance, and relative speed of vehicles;

Outperforms four major competitors on

accuracy, according to Axon testing claims;

Calibrated to register a scan only if the system

is at least 90% confident that it has identified all
of the license plate characters correctly;

Not an infrared camera, so will be able to read temporary

plates and plates that are not retro-reflective;

It likely will be able to recognize other vehicle

characteristics, in addition to extracting plate data.

15
III
Preliminary
Assessment
and Guidance
Regarding ALPR
Benefits & Costs

16
 We begin with what must be obvious: Axon’s
entry into this market with what appears
to be a powerful and low-cost ALPR has the
potential to increase dramatically the use of
ALPRs by law enforcement.
In our First Report (page 17), we wrote, “We data available to track vehicle movement.
have found it essential to be involved at the Axon’s ALPR-equipped cars will be able to
earliest possible stage of product development, scan three lanes of traffic simultaneously,
preferably before Axon even begins to design potentially front and back, reading and storing
a product. Early involvement means that we all the license plates those cars come across.
can guide Axon’s research and that we can There is publicly available information
suggest design modifications before they suggesting Axon is not the only vendor
become cost prohibitive for the company.” contemplating this course of action.
Axon has followed that recommendation and Thus, whether or not Axon pursues this
solicited our input at this developmental stage. course, it is likely that the state of affairs
In this section, we offer a preliminary we just described will come to pass.18
assessment of potential benefits and potential
We cannot and will not evaluate Axon’s
harms of low-cost ALPR systems in use by law
ultimate decision to enter the ALPR market until
enforcement. We do so using the benefit-cost
we have more specific information about its
Framework outlined in our First Report, which
ALPR. At that time, we anticipate knowing not
we have included here as Attachment B.
only what Axon’s product can do, and how—
We begin with what must be obvious: Axon’s and whether it has developed the product
entry into this market with what appears to consistently with our recommendations—
be a powerful and low-cost ALPR has the but also a great deal more about the market
potential to increase dramatically the use of Axon enters. If Axon is the only market
ALPRs by law enforcement. It will become entrant of its size, that presents one scenario.
viable for law enforcement to outfit their entire If Axon enters a burgeoning market, with a
fleet with ALPRs. Increased usage is likely to product that is capable of leading the way to
expand substantially the amount of historical regulated and ethical usage, that is another.

17
Preliminary Assessment of even identify people in a vehicle. Some of
these capabilities amount to massive force
Potential ALPR Benefits multipliers, quickly automating what an officer
would need to do by hand or by calling in to
dispatch. Yet others represent entirely novel
We begin with the potential benefits to society tools previously unavailable to police.
and law enforcement. As we make clear in
our Framework (see Attachment B), if a A major difficulty we encounter is that for
technology does not yield concrete benefits, the most part we lack clear data on how
there is no reason to proceed further. useful ALPRs presently are to these various
purposes. For almost every benefit of
As set out above, it is clear ALPRs are being ALPRs anyone advances, there is a dearth of
used for a variety of purposes, from locating evidence to make clear precisely how real
stolen vehicles to identifying vehicles or extensive that benefit is. We believe it
associated with individuals with outstanding is imperative that such data be developed,
warrants. ALPRs also have the potential so that society has the information it
to identify make and model of a car, or needs to regulate ALPRs appropriately.

1. IMPROVED REAL-TIME LOCATION

OF HOT LIST LICENSE PLATES
Law enforcement agencies frequently seek
to locate particular vehicles in real time. For
example, a license plate may be radioed
A major difficulty out to officers as part of a BOLO (“be on
the lookout”). Agencies can issue a BOLO
we encounter is that for almost any reason—from locating a
for the most part we stolen vehicle, to identifying one associated
lack clear data on with an Amber or Silver Alert. The severity
of the matter under investigation can vary
how useful ALPRs considerably. For example, the outstanding
presently are to these warrant at issue could be for a violent crime,
various purposes. For unpaid fines, or suspended driver’s license.

almost every benefit Before ALPRs, the best an officer could do

either was to remember a license plate of
of ALPRs anyone interest or search dozens of plates over the
advances, there is a course of their shift. This was labor intensive
dearth of evidence to and unlikely to be very successful. ALPRs make
a substantial difference here. Once license
make clear precisely plates of interest are compiled into a hot list, an
how real or extensive agency can use ALPRs to automate the search
that benefit is. process. Hundreds or thousands of plates can
be scanned by one patrol car and compared

18
against the hot list. In the event of a match, 2. IMPROVED HISTORICAL
the officer is notified and can investigate.
INVESTIGATIONS
Unsurprisingly, then, there have been a
number of studies that indicate ALPRs improve In addition to real-time alerts, many agencies
an officer’s ability to detect vehicle-related use ALPRs to assist with criminal investigations.
violations.19 With respect to stolen vehicles, ALPRs generate huge searchable location-
specifically, one study put the increase at history databases. For example, an officer
140%.20 More generally, there is substantial who knows where and roughly when a crime
evidence that officers with ALPRs will recover occurred can generate a list of vehicles
a greater number of vehicles, seize more near the scene at the time. This list might
contraband, and make more arrests.21 identify potential suspects or witnesses.
Or, investigators can examine the historical
Although these increases appear real when path of a vehicle, helping to build a case that
compared to officers without ALPRs, it is the vehicle was involved in an offense.
worth noting the data indicates that ALPR
hit rates remain astronomically low. Most Although there are reports of ALPRs being
importantly, they are lower still for serious used successfully for investigations, there
offenses. For example, the ACLU examined is little data or study about the frequency or
data from Maryland’s state data fusion effectiveness of this type of ALPR use.23 There
center, which aggregates plate reads from are some indications that agencies use ALPRs
three-quarters of Maryland law enforcement for general investigative purposes more
agencies. Of the over 29 million plate reads frequently than they initially expected, but it
gathered in the center’s data in the first half is impossible to assess the utility of ALPRs in
of 2012, only 0.2% were hits of any kind, this regard without knowing how often this
and of those, 97% were for a suspended or type of investigation occurs, for what sorts of
revoked registration or emissions violation. cases, and how often they are successful.24
The ACLU found similarly low hit rates for There is a particularly critical need for study
cities and agencies across the country.22 on how far back in time stored ALPR data

There is a particularly critical

need for study on how far
back in time stored ALPR data
proves valuable, and with what
frequency. Knowing this would
inform a decision about how
long to retain ALPR data.

19
proves valuable, and with what frequency. of this research focuses on mobile ALPRs. In
Knowing this would inform a decision about short, the suggestion that mobile ALPRs will
how long to retain ALPR data. At present, improve traffic safety, remains conjectural.
we know that there are several jurisdictions
that manage to conduct their investigations
with 30 days or less of data storage, 5. INCREASED REVENUES
suggesting that the usefulness of stored
ALPR use may increase revenue for a
ALPR data may diminish rather quickly.25
jurisdiction. We are loath to call this a benefit,
especially when it comes to represent
a significant portion of a jurisdiction’s
3. GENERAL CRIME DETERRENCE budget. We present the evidence in this
There is little evidence that increased regard in our discussion of costs. But some
arrest rates due to ALPRs decrease jurisdictions may consider it a benefit.
crime overall. The few studies examining
this issue suggest otherwise.26
6. REDUCING ARBITRARINESS
AND BIAS
4. IMPROVED TRAFFIC SAFETY
One intriguing potential benefit of ALPRs is
ALPRs in police patrol cars might be useful that they might be used to make policing less
in enhancing traffic safety, but again we have biased, less arbitrary, and more productive.
no evidence in this regard. There is some For example, it is well-documented that
evidence that stationary ALPRs in red light traffic stops often occur for pretextual
cameras—not mobile ALPRs—provide reasons, i.e., enforcing the traffic violation
some benefit by reducing fatal collisions.27 is simply a means of investigating more
More generally there is some research serious crime. What we do know is that hit
suggesting a negative correlation between rates on such stops are notoriously low,
receiving a moving violation and subsequent that unnecessary traffic stops interfere with
involvement in a fatal crash, though it is not people’s liberty, and that such pretextual stops
evident to what extent mobile ALPRs are disproportionately burden racial minorities.29
used to enforce moving violations.28 None

20
One could imagine how the use of ALPRs significant contributing factor. Axon has
might reduce the costs of traffic stops, because suggested to us that some portion of
ALPR alerts might be more indicative of these accidents are caused by officers who
vehicles that actually justify officer attention. are distracted by typing in (or “running”)
Particularly if agencies calibrate their hot lists license plates manually (as opposed to
to include only relatively serious reasons for other distractions, including cell phones
stops (as defined by the communities those and radios). If this is true, it is at least
agencies serve), agencies might be able to possible that ALPRs would minimize some
focus their officers on fewer, higher quality amount of this distracted driving, thereby
stops, rather than the somewhat arbitrary reducing officer-involved accidents.
manner in which stops are conducted today. We have two concerns about this benefit.
The Board certainly hopes that this positive First, although some figures show
use of ALPRs comes to be—i.e., rather than “inattentive/careless driving” as a significant
increasing enforcement of low-level offenses, (but not leading) cause of officer collisions,
agencies use ALPRs to focus their officers’ we have no evidence of the degree to
attention on high-quality stops; and that rather which running license plates contributes
than allowing officers unlimited discretion to to these incidents.30 And perhaps more
conduct stops, that ALPR use fosters attention important, as discussed below, we are
on an objective reason to stop a vehicle. uncertain whether officers should be running
license plates as often as some do.
We have not, however, seen any evidence
to date that the use of ALPRs reduces
stops for other reasons. Thus, this potential
benefit is speculative at present. Still, as we 8. PREDICTIVE ANALYSIS
discuss below, we believe there are design It is possible that ALPR scans could be used
steps that Axon could take to make this as an input into predictive analysis tools.
scenario somewhat more likely to occur. For example, ALPR data might allow police
to anticipate where a particular wanted
vehicle will be in the future, or it might
7. OFFICER SAFETY WHILE DRIVING provide a basis for identifying vehicles
operating in suspicious patterns.31 The
Officers are involved in vehicle accidents
Board has seen no evidence that such tools
at twice the rate of the average driver,
are near operational. Any such tool would
with distracted driving being a relatively
require its own careful ethical analysis.

In sum, it is possible to think of benefits of ALPRs, and some,

such as locating stolen vehicles or investigation of crime, may
be quite real. But the evidence in support of these benefits is,
at present, remarkably lacking, making it difficult to assess the
impact of the pervasive use of low-cost mobile ALPR units.

21
Preliminary Assessment of work more difficult around serious crimes.33

Potential ALPR Costs

Thus, the costs associated with ALPRs will vary
dramatically depending on how hot lists are
constructed and what prompts alerts sent to
After considering potential benefits, our officers. ALPRs obviously are going to alert
Framework turns to potential costs. As we officers most frequently to lower-level matters,
explained in our First Report, it is essential at unless they are set to disregard them.34
this step to try and capture not only obvious What’s more, the impact of increased
costs, but unintended consequences as well. enforcement will not be felt equally
As with benefits, many of these costs are across all communities. Typically,
difficult to quantify. They are real, but it communities of color and lower income
is hard to know how extensive they are, communities bear the brunt of increased
both because they have not been studied enforcement, for two related reasons:
sufficiently, and because ALPR use is going to First, mobile ALPRs go where officers go.
increase as the costs of mobile units drop. More officers are deployed to neighborhoods
that are perceived to be “high crime,”
which tend to be minority and low-income
1. INCREASED ENFORCEMENT neighborhoods. ALPRs thus will be more
active in those neighborhoods, both enforcing
GENERALLY AND ITS LIKELY real-time alerts and gathering historical
DISPARATE RACIAL, ETHNIC, AND ALPR data. This disparate impact was seen
SOCIOECONOMIC IMPACTS in a study from the Electronic Frontier
Foundation, which examined ALPR scans
At their core, ALPRs promote greater
in Oakland, California and found that areas
enforcement.32 That’s their purpose. If it
with greater numbers of black and Latinx
were not for their capacity to alert officers to
residents were more likely to be scanned.35
vehicles deemed by law enforcement worthy
of attention, the value of ALPRs would be Second, hot lists inevitably will reflect existing
severely limited in law enforcement’s eyes. racial, ethnic, and socioeconomic disparities
in the criminal justice system and our society
Every enforcement action is an intrusion on
at large. For example, to the extent ALPRs are
a person’s daily life. Those intrusions have
used to enforce expired vehicle registration
inherent costs—both for the individuals
or unpaid fines and fees, they will fall
stopped and for communities bearing the
disproportionately on low-income communities
brunt of enforcement efforts—that must
and communities of color.36 To the extent
be accounted for and balanced against the
ALPRs are used to enforce open warrants for
reason for the actions. This is especially true
low-level offenses (e.g., drug possession),
when it comes to over-enforcement of low-
they will reflect pre-existing disparities in
level offenses, which can have a corrosive
low-level enforcement across the system.37
effect on community trust in law enforcement.
This impact on trust and legitimacy may In short, ALPR technology necessarily operates
lead to people being unwilling to report within the context of police practices and
crimes or cooperate with law enforcement, deployment decisions that often have clear
which in turn makes law enforcement’s disparate racial and socioeconomic impacts.38

22
2. INCREASED FINES-AND-FEES has offered free ALPRs to law enforcement
agencies so long as those agencies hot list
ENFORCEMENT AND ITS DISPARATE plates of individuals with outstanding court
IMPACTS costs and fines. When these drivers are
pulled over by law enforcement, they are
Closely related to increased enforcement
told they can either be arrested or pay their
are the costs associated with increased fines
fine on the spot, with an additional 25%
and fees. As noted above, ALPR hit rates
“processing fee” paid to the vendor.44
for low-level offenses are far higher than
for serious crimes.39 Given the evidence Jurisdictions ultimately have to make their own
that ALPRs today sometimes are used to decisions regarding how ALPR enforcement
generate revenue, we are concerned that occurs, but the Board has real concern about
more ALPRs will mean increased collection imposition of these sorts of costs, which
of fines and fees by jurisdictions looking to potentially could increase notably as the
raise revenue. One major vendor has called costs of mobile ALPR use come down.
ALPRs “a proven force multiplier that helps
generate revenue,” and said that ALPRs
helped generate $200,000 in 30 days for one 3. INAPPROPRIATE STOPS
department, with “the end result [being] fewer
Increased ALPR use carries with it the risk
scofflaws getting away with nonpayment.”40
of unconstitutional, wrongful, or otherwise
Heavy fines and fees enforcement can have inappropriate stops. Such interactions pose a
severe impacts on individuals, often with danger if an officer stops a vehicle believing that
disparate racial and socioeconomic results.41 the individual in the vehicle is a violent offender.
Overreliance on fines and fees already One source of such erroneous stops is the
is a problem plaguing many areas of the ALPR technology itself.45 Inaccurate reads can
country. Smaller jurisdictions, particularly create false alerts that lead to unnecessary
those with high rates of poverty, frequently police interactions. Although this has
rely on fines and fees to fund huge portions occurred, we lack data on how often.46
of their budget. In a recent survey of 600
A different form of false positive occurs when
jurisdictions by GOVERNING magazine, at
hot lists are inaccurate or not up to date.
least 284 financed over 20% of their budget
Although we have no idea how extensive
from fines.42 That level exceeds the limit
these incidents are, there have been reported
set by the State of Missouri after the U.S.
instances of officers holding innocent people
Department of Justice investigation into fines-
at gunpoint because of erroneous ALPR
and-fees enforcement in Ferguson, Missouri,
alerts.47 There is evidence that outstanding
which found that the city’s law enforcement
warrants—a potentially large part of hot
practices were “shaped by [its] focus on
lists—include a variety of errors that could
revenue rather than by public safety needs.”43
lead to false ALPR alerts and enforcement
The more private companies align their actions against innocent drivers.48
financial interests with fines-and-fees
Finally, even if the technology works properly, it
enforcement, the more likely the disparate
goes without saying that a car is not the same as
impacts of such enforcement will be
a person. For offenses not directly related to a
realized. There are reports that one vendor
vehicle, an officer often stops a car because they

23
are interested in the person they believe to be over years.50 Some ALPR manufacturers
driving the car (usually the registered owner). own and aggregate ALPR scans, and share
But particularly in less affluent households, this data with purchasers, whether law
there are likely to be multiple drivers sharing enforcement or private entities.51 It also
a single vehicle. What’s more, the car-sharing may be possible to use ALPR data to predict
market continues to grow.49 Thus, it is a matter where vehicles can be found in the future.
of probability, not certainty, whether the As the Supreme Court has made clear,
person driving the car is the registered owner. location history can paint a detailed picture
of our personal lives. The longer the data is
stored, the more intimate the picture.52 For this
4. THE PRIVACY IMPLICATIONS OF reason, the Supreme Court has held that even
GROWING ALPR DATABASES though law enforcement viewing information
such as location is not a “search,” if historical
ALPRs raise very serious privacy concerns.
data over a certain period of time is searched,
They reveal information about drivers
the Fourth Amendment is implicated, and
in the vicinity of the ALPR systems, and
access to that data requires a warrant.53
amass an enormous amount of historical
data that could be used to determine There is significant concern in the public
an individual’s whereabouts. sphere at the moment about the use of face
recognition, for these very reasons. Yet,
We begin by noting that it is a reasonable
while the widespread use of face recognition
question why law enforcement officers are
is still hypothetical in this country, the use
permitted to scan license plates at random
of ALPRs is ongoing and the data is readily
on the road at all—whether by manually
available. We do not know enough about
running plates or via ALPR. Apparently, the
the use of this ALPR data, but the privacy
practice is widespread and routine. It may well
risks here are evident and substantial.
be that because license plates are displayed
publicly, scanning and running plates is not Additionally, the collection of ALPR data
a “search” within the meaning of the Fourth runs the risk of chilling First Amendment
Amendment, under existing precedent. That protected activities, such as attending houses
does not mean, however, that such scanning of worship or public protests. There already
necessarily is appropriate. To be clear, officers is evidence that ALPRs have been used for
may well have sufficient legal cause at times investigations around First Amendment
to want to know more about a particular protected activity, both by law enforcement
automobile. But it is not clear why individuals and groups with a particular political agenda.54
going about their lives and business, and There also is a very real risk that ALPR data
doing nothing that rises to the level of will be abused by those with access. There
reasonable suspicion, should have anything have been documented cases of officers
about themselves investigated or recorded. using ALPR data to stalk or harass individuals,
The massive ALPR databases being or to tamper with and sell records they
amassed allow those with access to obtained, though again we have no way
reconstruct a history of where a car has of knowing how widespread this is.55
been. Using one of these databases, a Finally, there is the risk of data breaches,
reporter was able to track a man’s vehicle which can lead to personal data being used

24
for any number of troubling purposes. For susceptible to generate ALPR alerts for minor
example, a 2015 investigation found more than offenses and fines-and-fees enforcement.
100 exposed ALPR systems online, “often with The NCIC Vehicle List generally is restricted
totally open web pages accessible by anyone to serious crimes. But NCIC also includes
with a browser.”56 Three years later, a similar civil immigration violations (such as visa
investigation found that little had improved, overstays).60 It therefore is possible that when
with the information from more than 150 running names or plate through NCIC, an
ALPR cameras searchable on the internet, alert will be generated solely for immigration
most using factory-default passwords.57 enforcement. The Board is deeply troubled
by the possibility that the spread of ALPRs
might co-opt local law enforcement into
5. ADDITIONAL POTENTIAL COSTS federal immigration enforcement when those
agencies otherwise would not choose to
DEPENDING ON USE engage in such enforcement.
There are additional potential costs of ALPRs,
depending on how an agency chooses to use
Expanded Capabilities Over Time:
the technology. We cannot possibly know or
review every scenario here, but we note some At the time that Axon has a final product
additional major concerns that arose during available for review, we will ask precisely
our discussions: what data or images the ALPR can collect
and analyze (e.g., license plate, state of
origin, type of vehicle), whether the ALPR
Gang Databases:
camera can be repurposed to capture
The Board would have huge concerns additional or different data (such as through
about agencies linking their ALPRs to gang use as a stationary ALPR), and whether
databases in either direction—i.e. to use the software can be run on other cameras
ALPRs to alert officers when they are in the (such as CCTV or body-worn cameras). This
vicinity of a vehicle believed to be associated information will allow us to think about
with gang activity, or to add individuals specific use cases, and say more about them.
or vehicles to a gang database based on
But in addition to specifics about how an
ALPR scans. There are reports that this is
agency actually will use ALPRs, it is important
occurring in some jurisdictions.58 These sorts
to be conscious of how the technology can
of lists have been criticized as inaccurate
develop over time, and what capabilities
and overbroad, leading to the possibility of
could simply be added on to existing systems.
unnecessary enforcement, and perpetuating
For example, at present ALPRs do not read
racial disparities.59
vehicle registration stickers. Adding this
capability would raise serious questions
Immigration Enforcement: about increasing enforcement of a violation
The Board has a particular concern relating to that, at times, simply reflects financial
the federal NCIC system and its relationship hardship. Additional capabilities—such
to immigration enforcement. In general, the as capturing non-plate text (e.g., bumper
Board favors using the NCIC system over stickers) or photos of passengers—would
local hot lists, as local hot lists are more raise a number of serious considerations. The
same is true of any predictive capabilities.

25
6. LACK OF INFORMATION AND from publicizing their use of ALPRs or
disclosing their agreements to the public.61
TRANSPARENCY AROUND ALPR USE Lack of transparency around how and where
There is too little information about the ALPRs are being used magnifies many
details of how ALPRs are used across of the costs discussed above, as well as
the country, including where ALPRs limiting information about benefits. When
are deployed, the compositions of hot law enforcement agencies operate without
lists (both what types of crime, but also transparency, it detracts from their legitimacy,
demographic breakdowns), how often and inhibits their ability to act effectively. And
historical data is accessed, and much more. communities cannot make decisions about
Vendors must share some of the responsibility ALPR use without information. Although some
for this lack of transparency. There are degree of secrecy around law enforcement is
provisions in certain contracts that actively essential at some times, the Board is skeptical
impede transparency, barring agencies that is the case around the use of ALPRs.

In summary, with respect to both benefits and costs, we

are operating with limited information, often extrapolating
from anecdotes, and doing our best to estimate effects.
We have acknowledged there are potential benefits, but
it is clear to us that there are very real costs as well.

26
27
IV
Recommendations

28
 What follows are a set of recommendations
that we hope will be of use to society as a whole,
including legislative bodies, communities
concerned about policing, law enforcement
agencies, and the broader ALPR industry.

We have identified what we know Too often, in law enforcement, as in other

regarding benefits and costs, as well as parts of society, we operate from the
the product Axon anticipates producing. assumption that because it is possible to
The recommendations that follow provide do something, we should. Technology
guidance on what we believe an ethical and has exacerbated this type of thinking. As
appropriate usage of ALPRs might look like. explained in our First Report (pages 31–32),
Although we understand that ALPRs are used we take the view that one antidote to this
internationally, our recommendations focus problem is that decisions around the use of
on the United States, and are sensitive to powerful surveillance tools “should be made
U.S.-specific laws and policy challenges. in an open and transparent manner, in ways
that allow for democratic accountability.”
In light of the concerns we have expressed
We think it essential that the public debate
in this report, we do not limit our
occur before law enforcement acquires and
recommendations to Axon. Rather, what
uses such tools, not after we have become
follows are a set of recommendations that
acculturated to them. This is particularly
we hope will be of use to society as a whole,
important so that law enforcement does not
including legislative bodies, communities
move ahead of public sentiment, only to run
concerned about policing, law enforcement
into public concern that undercuts legitimacy
agencies, and the broader ALPR industry.
and trust in the public safety mission.
We believe that government

01 Law enforcement agencies

should not acquire or use
regulation of ALPRs is imperative.
As detailed in Part II, ALPRs in the U.S. have
ALPRs without going through an been deployed widely, with few limits. In
open, transparent, democratic the United Kingdom, however, ALPRs are
used widely, but within a detailed regulatory
process, with adequate opportunity
and policy framework. Protections include
for genuinely representative public publishing information about how ALPRs are
analysis, input, and objection. To used, information about data-retention periods,
the extent jurisdictions permit ALPR and who is authorized to search historical
use, they should adopt regulations databases.62 We do not endorse any particular
part of that regulatory scheme, but believe
that govern such use. (This is what
there should be similarly comprehensive
we said about face recognition, ALPR regulation in the United States.
and it is true as well for ALPRs.)
Courts are unlikely to offer this regulation.

29
United States constitutional law is lagging statute’s exception for criminal investigations
behind the technological revolution we are and criminal intelligence gathering.69
undergoing. It excludes many law enforcement
Most state regulations, however, include
activities from the scope of constitutional
only general language requiring a law
review simply because they occur out in
enforcement or public safety purpose, and
public, such as taking pictures of our faces or
do little to limit how ALPRs can be used.70
our license plates. But such pictures are used
to gain more information and may well lead In our view, although ALPR use is local and
to enforcement or other social costs. Simply should be regulated at the lowest level
because technology makes it possible to do possible—so that each community can
something at scale does not mean we should express its view of how local law enforcement
not ask the question why government should and private companies will operate—the
be able to collect such information without nature of cross-jurisdictional databases
some specific basis. Constitutional law presents suggests a need for federal legislation as
a floor, whereas legislation can go further. well. Technology vendors, law enforcement
agencies, civil society, community groups, and
At present, the regulation that exists in the
legislators should work together to achieve
U.S. is piecemeal and does little to tackle
meaningful regulation of ALPRs. Legislation
the systemic risks. There is no federal law
should address, at a minimum, the issues
governing ALPRs.63 The federal Driver’s Privacy
we have raised in this report, including the
Protection Act (DPPA) provides some protection recommendations that follow.
to Department of Motor Vehicle records, but
does not limit law enforcement investigations,
nor does it impact the ability of private
vendors to amass billions of ALPR reads.64 02 Agencies should not deploy
ALPRs without a clear
Several states have adopted regulations use policy. That policy should
around ALPRs. Most do so by statute, though be made public and should,
a few executive agencies also have issued
at a minimum, address the
guidance.65 New Hampshire, for example,
appears to have adopted what we call a concerns raised in this report.
“Snapchat model”—barring stored non- This recommendation requires little
evidentiary plate reads for any period of explanation. We believe both that agencies
time.66 A few other states include relatively should not deploy surveillance technologies
short limits—from 21 days to 90 days.67 But without clear policies governing their use,
many others permit retention for years.68 In and that basic transparency requires that
Virginia, the Supreme Court, confronting what these use policies be made public. It is
it called “sweeping randomized surveillance possible some operational aspects will
and collection of personal information” via need to be kept confidential, but we expect
ALPRs, has held that images taken of license that to be the exception, not the rule.
plates are protected personal information Unfortunately, this is far from the status
under the state’s 1976 Government Data quo. A few states require law enforcement
Collection & Dissemination Practices Act, agencies to have written ALPR use policies.71
and rejected law enforcement’s argument A subset of those, such as California,
that this type of surveillance fell under the require that those policies be posted on

30
agency websites.72 In our view, however, the settings they will use on their ALPRs (such
whether required by state law or not, ALPR as data retention periods and the criteria that
policies should be disclosed publicly. trigger alerts), and data from ALPR use such
Axon has suggested to us that it could as heat maps and hit rates. This website could
create a website or web tools for each law be integrated with Axon’s ALPR such that
enforcement agency that purchases its ALPR usage statistics were automatically uploaded,
system, where all of these policies—and without requiring any additional work on
much more information—could be made the part of the law enforcement agency.
available to the public. We believe this is The Board is well aware that this sort of
the right direction, as we explain in greater transparency is not the norm in the policing
detail below. We believe transparency tech industry. Indeed, vendors often require
around the use of ALPRs is essential. secrecy from law enforcement, and at the same
time sometimes impose requirements that give
the vendors nearly unlimited rights to use and

03 Vendors, including Axon,

should design ALPRs to
share data.74 We cannot accept this status quo.
Axon not only should design for transparency
facilitate transparency about their in the way suggested above, but should
use, including by incorporating incentivize agencies to use the product’s
transparency features. For example, Axon
easy ways for agencies to share
should include these web tools as a default
aggregate and de-identified data. part of its ALPR package, and offer monetary
Each agency then should share this discounts to agencies that leave it enabled.
data with the community it serves. Communities should be able to learn easily
if their agency is being transparent. Axon
Throughout this report we have discussed
should also, in consultation with the Board, law
potential costs of ALPRs, such as the privacy
enforcement, and community stakeholders,
impact of long retention periods for data,
develop a model ALPR use policy, make
or the fact that ALPRs can be used for low-
it widely available, and offer resources to
level enforcement with disparate burdens
customers to help customize the policy if they
on communities of color or economically
agree to make it public.
disadvantaged communities. Some of these
problems can be mitigated to some extent
through transparency. At the very least,
communities should know how long ALPR data 04 Vendors, including Axon,
should design their ALPRs
is being retained, and how ALPRs are being
used, i.e. for serious felonies alone, or for low- so that agencies can adjust the list
level or traffic offenses. “Heat maps” can show of vehicles to which an ALPR will
the neighborhoods where ALPRs are employed alert law enforcement officers, so
most frequently.73 Hit rates can show the that the list includes only those
efficacy of ALPRs for various items on a hot list.
offenses or reasons most of concern
Axon has represented that it has the capability to that agency and its community.
to build a “transparency portal”—a publicly
Although communities must
accessible website, customizable for each
community—where agencies can post policies, decide the contours of their own

31
alert lists, as a general matter we ALPR scans in particular neighborhoods, and
believe that these lists should not periodically seek to correct any imbalances (for
example, by limiting the number of scans that
be used to enforce civil infractions, occur in already saturated neighborhoods or
offenses enforceable by citations, or by alerting only for the most serious offenses).
outstanding warrants arising from Doing so would help to ensure that simply
a failure to pay fines and fees. because a person lives in a “high crime”
neighborhood, they will not necessarily be
As detailed above, ALPRs can be used solely for
subject to disproportionally more surveillance
very serious matters, such as stolen vehicles,
or enforcement for less serious matters.
Amber and Silver alerts, and to identify
vehicles associated with violent felonies. We commend Axon for considering these
Conversely, ALPRs could be used much more controls, and encourage that its ALPR be
widely, such as for traffic or fines-and-fees built with them integrated and available for
enforcement. Although we feel strongly that agency use.
using ALPRs for low-level offenses and fines-

05
and-fees enforcement is not appropriate and
Vendors, including Axon,
will have serious deleterious consequences,
ultimately, communities should confront must provide the option to
and address these issues for themselves. turn off immigration-related alerts
Axon has said it can build its ALPR so that from the National Crime Information
communities and agencies can choose Center so that jurisdictions that
specifically the matters for which officers choose not to participate in federal
will receive alerts (for example, permitting immigration enforcement can do so.
alerts for outstanding felony warrants
and stolen vehicles, but not overdue fines This recommendation is closely related to
or civil infractions). Axon should design the previous one. Although some jurisdictions
its technology with default settings to have chosen to cooperate with federal officials
encourage ALPR use in ways that avoid some on immigration enforcement, others have
of the more serious concerns expressed expressed the view that they do not wish to be
here, including over-enforcement of civil part of the federal government’s immigration
infractions, offenses enforceable by citations, enforcement efforts. Although it is complicated
or outstanding warrants arising from a failure issue, as a general matter, the law is clear
to pay fines and fees. These same default that the federal government cannot force
limits should be applied when a jurisdiction local law enforcement agencies to enforce
links to external hot lists, such as NCIC, the immigration laws.75 The Board believes
so that the ALPRs automatically filter out strongly that ALPRs should be designed in
offenses that do not meet these criteria. a way that preserves this local choice.

Axon also has told us that it can design The complication here is that the federal NCIC
features into its ALPR that track the geographic dataset includes data on both criminal and
disparities that ALPRs produce, given that civil immigration violations. Immigration-
officers are not deployed evenly across a city. related files can include license plates,
For example, Axon’s technology could include and running an individual’s name through
features that assess the concentration of NCIC also can trigger immigration alerts.

32
In our view, communities must confront this
issue for themselves and should not be forced 07 An ALPR alert, on its own,
should not constitute sufficient
into another path simply as a function of the
technology they purchase. If ALPRs are by
grounds to stop a vehicle. Officers
default alerting for immigration offenses, must make visual confirmation
individual officers might make enforcement independently that the license plate
choices based on the alert even if contrary to matches the hot-listed plate. If the
community preferences. Therefore, in order to offense at issue is associated with
preserve each community’s ability to choose,
vendors—including Axon—should design
the registered owner of the vehicle
their technology such that immigration-related (as opposed to the vehicle itself), the
offenses in NCIC do not automatically trigger officer also should ascertain whether
ALPR alerts. the driver is consistent with the
description of the registered owner.

06 ALPRs must be designed and

operated in ways that ensure
Because of the potentially harsh consequence
of false positives, we do not believe it is
alert lists are checked routinely sufficient for an officer to receive an alert and
respond, or even simply to be told it is her
for errors and kept up to date.
or his job to confirm that the license plate
Based on our preliminary conversations with of a vehicle matches the alert. The Board
Axon, we believe that ALPRs can be designed has discussed with Axon various ways to
to minimize stale and inaccurate data. For minimize the impacts of false positive alerts
example, Axon might incorporate features by implementing “human in the loop” design.
that require agencies or officers to confirm Although the final design is for Axon to decide,
periodically that a “vehicle of interest” remains we expect the design to ensure that officers
of interest, rather than letting license plates sit verify using the actual license plate in their view,
on a hot list indefinitely. Or Axon might allow not by reflexively approving the ALPR’s read.
for automatically removing hot-listed plates by
When the offense at issue relates to a person
cross-referencing hot lists against other datasets
(typically the registered owner) and not the
such as recently paid tickets or recently paid
vehicle, additional steps should be taken to
registration fees. The better option might be
confirm that the driver matches the description
to require hot listing to be repeated at certain
of the person that law enforcement is seeking
intervals, so that the ALPR could reject alerts
(typically the registered owner). Whether or
from hot lists that have not been updated
not this is required as a matter of constitutional
and confirmed within a certain time span.
law, this practice can be essential in avoiding
That said, we doubt this problem can be solved needless stops, particularly in communities
by technology alone. Agencies can help ensure in which vehicle sharing is common.77
ALPR data is accurate and up to date through
Officers should be trained that ALPR matches
their use policies. Agencies, for example, can
alone are insufficient and must be instructed
require a second person to double check every
on what to do when they receive an alert, and
entry onto a hot list, can institute periodic
agency policy should mirror these requirements
reviews of hot-listed plates, can routinely scrub
in a verifiable and auditable way.
stale warrants and paid fines, and so on.76

33
08 Axon should work with
partner agencies to determine
at all), we are not yet prepared to recommend
a specific length of time. With additional
information we expect to be able to make a
the shortest possible retention more specific recommendation in the future.
period for ALPR data that will
serve law enforcement needs
sufficiently (as explained below),
and set that period as the default
09 ALPR design should create
audit trails both of real-time
retention setting on its ALPRs. ALPR alerts and agency accessing
of historical ALPR data. Law and
At present, we have very little information
as to how often historical ALPR information agency policy should require
is used, how far back in time these searches regular auditing of ALPR usage.
go, and for what sorts of cases. For this We have outlined above the concerns with
reason, we do not, at this point, suggest a retaining and searching historical ALPR
specific data-retention limit. We suspect, data, as well as the potential for abuse and
though we cannot be certain, that even when misuse of ALPR systems. In light of these
agencies conduct investigations, they rarely concerns, we believe it essential that access
use very old data. Still, concrete answers to historical data be monitored closely.
to these questions would be invaluable for
sound policymaking. We encourage Axon to Vendors can facilitate this monitoring
work with partner agencies to study these through design. For example, vendors
issues and share their findings with us. can design their ALPRs so that supervisor
approval is required before any historical
Retention of ALPR data—that is, default data is accessed or searched. That might
retention of a plate scan absent a hot-list be limited to key supervisors, as is the case
match or some specific investigative need— (we understand) in the United Kingdom.78
for periods of time poses serious risks, as
outlined in Part III. Some agencies, by law, Communities also should consider whether
use a “Snapchat model” in which scanned their agencies must obtain warrants before
data that do not result in a hot-list match are historical scan data ever is accessed—a
not retained for any period of time. Some requirement Axon should make available on
agencies or states have short retention its technology. In any event, the technology
periods, others relatively long ones, and should assure warrants are utilized in
others have no stated end to retention. any situation in which governing law so
requires (for example, when seeking more
The Board believes that retention periods must than seven days of historical data).79
be as short as possible. Jurisdictions may want
to use ALPRs to identify stolen vehicles, help Design allowing for close monitoring of ALPR
with Amber and Silver alerts, and track down use applies after-the-fact as well. For example,
the most serious violent offenders. A “Snapchat ALPRs should be designed to generate clear
model” of ALPR use could accomplish much audit trails. At a minimum, these audit trails
of this with absolutely no data retention. should identify the officer that added a license
plate to a hot list, any instance when a hot-
Although we believe that there must be hard listed plate is identified in real time, and every
data retention limits (assuming data is retained instance when historical data is searched. The

34
audit trail should provide departments (or second. Axon can do this by making sharing
oversight bodies) with sufficient information data with private third parties more difficult,
to review both the agency’s and any officer’s and by being transparent about which law
actions for compliance with law and agency enforcement agencies chose to do so.
policy. Knowing these audit features exist
Per this recommendation, we encourage law
hopefully will dissuade misuse of the
enforcement agencies and communities to not
technology.
purchase access to privately owned ALPR data
that is conditioned on providing the private

10 Stored ALPR data must

be encrypted and secured
entity with access to the jurisdiction’s own
ALPR data.

against outside access and breach.

This recommendation should go without
saying, except there are many documented
12 ALPRs should be designed
such that if agencies
cases in which ALPR data has been left share data with other law
unencrypted and available to the world.80 enforcement agencies, they do
so transparently and in a way

11 ALPR vendors should

not retain the right to
that is governed by formal and
lawful data-sharing agreements.
access or share ALPR data, and Sharing historical ALPR data is a complicated
law enforcement’s ALPR data issue. As discussed in the previous
never should be shared for use recommendation, sharing with private parties
raises a host of potential concerns. But sharing
by for-profit third parties.
with other law enforcement agencies does so
It is one thing to authorize law enforcement to as well. For one thing, although one jurisdiction
conduct criminal investigations with ALPRs. It is or department may have regulations or policies
quite another thing for third-parties to benefit concerning their collection and storage of ALPR
financially from that ALPR data. Mixing law data (for example, the length of retention),
enforcement ALPR use with private financial these regulations can become moot when they
motives threatens to undermine legitimacy share the data with an agency that does not
and trust in law enforcement in much the same observe them. This is certainly the case when
way as over-enforcement of fines and fees. the data is shared into a private, unregulated
As such, the Board recommends that: (1) database such as those that operate today.81
private vendors, including Axon, not retain Because Axon intends to provide not only
ownership or the right to share ALPR data; and ALPR-enabled cameras but their database
(2) law enforcement refrain from sharing ALPR technology as well, it again can have a
data with private entities (e.g., repossession palliative impact on this state of affairs. First,
companies, insurers, private investigators). Axon should require agencies to make their
Axon can exert huge influence on these issues sharing arrangements transparent, or at a
by making a public pledge to abide by the minimum incentivize them to do so. Second,
former recommendation, and exhorting its Axon should design its sharing arrangements
law enforcement partners to abide by the (either contractually or technically) such

35
 It is essential that we know more so that we can properly
assess benefits and costs and set policy in a way that
allows for the former while minimizing the later.

that even once historical data is shared, it provide instructional materials in various
continues to follow the regulations of the forms to educate agencies and officers both
originating jurisdictions. This means that if as to why the defaults were chosen, and the
data originated in a locality with a 21-day consequences of choosing other settings. Axon
retention limit, it will retain that limit after it is has indicated it has the capability to do this.
shared. If the data originates in a jurisdiction
Also, as should be clear from the above, even
that bar sharing with private parties, that limit
in a well-working system, there is a lot of room
too should follow the data.
for error, such as items on hot lists that do not
belong there, misreads by ALPRs and the like.

13 Vendors, including Axon,

should never profit from
The costs of errors are substantial. Axon should
provide instructional materials as to best
practices for officers to deal intelligently with
fines and fees obtained through these situations. Axon has indicated it has the
law enforcement use of ALPRs. capability to do this.
When private companies profit from law
enforcement use of technology, it both
creates adverse consequences for individuals 15 It is imperative that data-
gathering and impartial study
and creates inappropriate financial incentives
for jurisdictions (particularly smaller ones) to be conducted of ALPR usage, so
over-engage in fines-and-fees enforcement. that communities and the country
This simply should not be allowed. are aware of how ALPRs are being
used, of what is required to make

14 Vendors, including Axon,

should provide adequate
that usage effective, of any harms
arising from ALPR usage, including
training materials for agencies and whether ALPRs are exacerbating
officers using its ALPRs, including disparities, and of ways to eliminate
about default settings and why or mitigate those harms.
they are set the way they are. As we have said throughout this report, we (and
We have discussed with Axon the concerns the country) are operating without sufficient
expressed in this report, and we also have information. Before the use of ALPR technology
discussed default settings on various aspects expands further, it is essential that we know
of ALPRs (such as for retention periods or more so that we can properly assess benefits
transparency requirements) to mitigate and costs and set policy in a way that allows
some of those concerns. Axon should for the former while minimizing the later.

36
37
V
Future Study
& Stakeholder
Engagement

38
What we’ve written here will not be our final word on this
matter. Both Axon and the Board must continue to learn and to
engage with community and law enforcement stakeholders.82
Below we outline a few preliminary ideas on next steps:

Continued Research additional study of our own between now and

product release, although that would require
both funding and law enforcement agency
It simply is impossible to weigh the benefits cooperation. Whether we are able to do so or
against the costs without information as not, the public would benefit from knowing:
to how valuable those benefits in fact are • The breadth of existing hot lists across
or how often the costs are encountered. a variety of jurisdictions, including the
We have had limited time thus far to proportion of various types of crimes
consider this issue. As such, we are on federal, state, and local hot lists;
confident that there is much we do not • Hit rates and efficacy for serious offenses;
know about ALPRs—there almost certainly • Hit rates for low-level offenses,
is research and thinking on ALPRs that we including traffic offenses;
have not yet come across. We intend to
• How traffic violations are and could
continue our review, and we encourage
be enforced using ALPRs;
anyone to share information with us.
• How officers experience ALPR use on
But still, based on what we do know, there patrol and how it impacts their behavior;
is a real dearth of concrete information
• How often there are false positives,
about ALPR use and a variety of unanswered
and for what reasons;
questions that might be critical to our
future evaluation of Axon’s ALPR. Board • The costs of false positives; and
members are considering conducting some • The value of retaining license-plate reads.

39
Axon ALPR Timeline

May/June 2019
Axon first raised potential ALPR product with the Board
and shared potential ethical-design elements.

Summer 2019
Subcommittee of Ethics Board met several times to
discuss ALPRs in advance of September meeting.

September 2019
Board met to discuss ALPRs and possible Board statement.

October 23, 2019

Axon announces its ALPR product, and Board releases report.

Q1 & Q2 2020
Study and stakeholder engagement, as outlined below.

Q4 2020
Updated assessment report from Board.

40
Stakeholder Engagement Model Statute & Model
ALPR Policy
Axon has expressed interest in hearing from
law enforcement, the advocacy community,
Learning from our additional study and
impacted communities, government officials,
stakeholder engagement, the Board
and other stakeholders on any issue related
intends to draft both a model ALPR
to ALPRs that those stakeholders deem
statute and a model ALPR use policy.
relevant. At present, we have a few ideas about
what form this engagement might take: The former will be aimed at communities and
government officials interested in bringing a
level of restraint to this otherwise unregulated
• As with any product development, Axon industry. We do not intend to lobby any
engages law enforcement agencies through particular jurisdiction to adopt our model, but
the development stage. The Board would we are happy to offer guidance in support
like to help inform that engagement and of those considering legislative action.
learn from it. We encourage Axon to focus
Our model use policy will be aimed primarily
on agencies of all different sizes during the
at law enforcement agencies. We expect to
course of this engagement.
go beyond a bare bones policy and address
some of the more difficult issues raised in
• The Board is open to feedback from this report, including conditions for adding
all stakeholders on the various issues a license plate to a hot list, obtaining and
raised in this report. We are discussing documenting supervisor approval before
possible mechanisms for obtaining this accessing historical data, public reporting
information, including paid focus groups around ALPR use, and much more. We
or an open comment period, but also understand that law enforcement agencies will
hope that organizations with ties to these want to customize this policy for their particular
communities will help us in this process. department, and we hope Axon will be able
to offer this type of support to its customers,
along with training around proper ALPR use.

41
VI
Conclusion

The Board sees the potential value in ALPRs, but we also find the current state of affairs
deeply concerning. Law enforcement and private actors vacuum up plate data across the
country creating databases of billions of scans, too often available to anyone willing to pay.
These scans are sometimes used to enforce civil infractions, collect fines and fees, or even
pursue purely private profits. There is evidence this enforcement falls disproportionately
on low-income individuals and communities of color, and much of this occurs out of the
view of the public and with minimal democratic input. As the purchasing costs of ALPR
technology drop, it only will become more widely available and more powerful.
In this report, the Axon AI Ethics Board hopes to call attention to this state of affairs with four
goals in mind: (1) that communities and governments will heed our call for democratic regulation
of this industry; (2) that Axon will take our guidance and recommendations into account as it
develops its ALPR; (3) that law enforcement agencies currently using ALPRs will modify their
policies to comply with our recommendations; and (4) that other vendors will follow suit.

42
43
VII
Attachments

44
Attachment A: 10 Questions ALPRs CAN AID LAW ENFORCEMENT:
to Ask Your Law Enforcement • ALPRs can help locate missing
children or adults through the use
Agency About ALPRs of Amber or Silver alerts.
• ALPRs can help track officers track down
more stolen cars by increasing the
WHAT ARE ALPRs AND WHY efficiency with which plates are checked.
DO THEY MATTER? • ALPRs can help locate people wanted
Automated License Plate Readers (“ALPRs” for investigation of serious crimes,
for short) are one of the most widely used including witnesses and suspects.
law enforcement systems in existence. • ALPRs may help remove individual officer
Although the adoption of ALPRs has been discretion from police stops, potentially
rapid and widespread, it has happened leading to more equitable law enforcement
with too little public oversight. and criminal justice outcomes.
ALPRs are camera systems that record license
plates, as well as the time and place of where
the license plate was scanned. ALPRs often
UNREGULATED USE OF ALPRs
are capable of taking in much more than CAN LEAD TO SERIOUS ISSUES:
license plates, including the surrounding scene, • ALPR errors can cause the police to stop
and even images of drivers and passengers. the wrong vehicle or wrong person.
ALPRs can be mounted almost anywhere: • ALPR-aided enforcement can worsen racial
light poles, overpasses, police cars, private and socioeconomic disparities, particularly
tow-trucks, or even run on smart phones. when used to enforce low-level offenses or
Many police departments use ALPRs in generate revenue through fines and fees.
patrol cars. The ALPR alerts officers if it • ALPR data can be used to reveal details about
scans a vehicle whose license plate is a person’s life, leading to privacy concerns.
in the system, perhaps because the car • ALPRs can be used in ways that chill
was reported stolen or is associated with First Amendment liberties (such as
someone with an outstanding warrant. monitoring all the cars parked outside a
Besides being used to alert law enforcement particular place of worship) and threaten
to vehicles of interest, the information they Fourth Amendment rights by allowing
collect can be stored away for future reference. law enforcement and private companies
Thus, if the police want to know where a to track your location over time.
vehicle has been at different times, they can Unfortunately, because of lack of data
consult the stored license-plate database. and study, we know very little about how
large the benefits or costs are—all the
more reason every community should
do its best to be informed about how
its law enforcement is using ALPRs.

45
10 QUESTIONS TO ASK YOUR LAW violations, outstanding fines and fees)?
• On what basis will the agency add a license
ENFORCEMENT AGENCY ABOUT ALPRs plate to its alert list (or “hot list”)?
1. Is your agency using Automated
7. What protections are in place to keep
License Plate Readers (ALPRs)?
ALPRs from being used in a way that has a
• How long have they been in use, disparate racial or socioeconomic impact?
and how many do they have?
• Where are ALPRs deployed, and does it result
• Are they mobile (on police cars) or in unequal surveillance of some communities?
fixed (on light poles, etc.)?
• What demographic data is being tracked to
2. Did the public or elected officials have allow the public to audit the use of ALPRs?
an opportunity to provide input before the
8. What types of data does your
agency acquired or deployed ALPRs?
agency retain from ALPRs?
• Who was consulted about the
• Does the agency keep only the records of
decision to purchase ALPRs?
license plate characters, or does it retain the
• Did the agency weigh the benefits and costs of video or other related images, include the
using ALPRs before purchasing the equipment? surrounding area and pictures of the drivers?
3. Has your state or municipality passed • Is the data cross-referenced against
any laws governing the use of ALPRs? other databases (e.g., gang or
• Do these laws regulate use, transparency, criminal history databases)?
data-retention limits, reporting or audits? 9. How long does your agency
• Is your agency in compliance retain ALPR data?
with all applicable laws? • Does the agency keep scan data for
4. Where did your agency acquire the 7 days, 30 days, 90 days, a year, etc.?
ALPRs it is using, and under what terms? Or does your agency scan plates for
• Are the contracts publicly available? live alerts only, without retaining the
information, absent evidence of a crime?
• Is there a non-disclosure agreement barring
public disclosure of certain ALPR functions? • Do different rules govern your agency’s
use of ALPR data when accessing historical
• Who owns and retains rights to the ALPR data?
data beyond a certain period of time, such
5. Does the agency have an ALPR use as requiring a warrant to run a search?
policy? Is it publicly available?
10. Does your agency share data with or
• Who is authorized to use ALPRs access data from any other law enforcement
and under what conditions? agency or private party? Under what terms?
• Do certain functions require supervisory • Does your agency share its ALPR scans
approval, such as searches of historical data? with, or incorporate scans from, private
6. What types of offenses does the companies (e.g., tow trucks, private
agency enforce using ALPRs? investigators, debt collectors, etc.)?
• Is ALPR-use limited to serious crimes (e.g., • If so, under what agreements, and
auto theft, Amber alerts) or are they used for how are outside parties accountable
low-level offenses (e.g., traffic, immigration for handling your data?

46
Attachment B: Product Evaluation Framework

Step 1 Step 2 Step 3 Step 4

Describe the use Begin to evaluate Begin to evaluate Consider

case at issue, but potential benefits, potential costs, product design
do not attempt to being careful to being careful to and features
define all the details try and capture try and capture to maximize
of product design. unintended unintended benefits while
impacts. consequences. minimizing costs.

GUIDANCE ON ASSESSING 3. How certain is it that the technology

will address the problem?
POTENTIAL BENEFITS
• Have there been evaluations
Although most products bring an expected (either internal or external)?
benefit, we found it essential not to speed
• Are there product performance concerns
through this stage of the process. Thus, for each
that might limit effectiveness?
use case, we asked:
• Will benefits be evenly distributed
throughout society?
1. What is the specific problem(s) the
• What countermeasures might individuals
product is intended to solve?
take in response to the adoption of this tool,
• “Problem,” here, might be a law enforcement and how much would such countermeasures
problem (e.g., improving law enforcement reduce the expected benefits?
methods), it might be a social problem, or it
might be a problem relating to the internal
operations of a police department. It is 4. Could using the technology have
important, particularly when framing the unintended or secondary benefits
problem as a “law enforcement” problem, on any of the following issues:
to be able to articulate the public safety goal • Minimize criminalization of low-level offenses?
that would be addressed through the use • Additional control and protection
of technology, rather than considering “law of personal data?
enforcement needs” as an end in itself. This
• Mitigation of racial and/or identity bias?
approach is essential to ensuring that the
ultimate consumer of every Axon product is • Improved transparency or public trust?
the community that a policing agency serves. • Better compliance with U.S.
constitutional requirements?
• Other societal benefits?
2. How important/what is the magnitude
of the problem you expect to solve?

47
GUIDANCE ON ASSESSING statistically unrepresentative training data
or exhibiting any algorithmic bias)?
POTENTIAL COSTS • Disparities in operation (e.g., whether
We have done our best to maintain a holistic the technology might be used in
assessment of potential social costs of a ways that create or exacerbate
given technological use.1 We are well aware identity bias and/or disparities)?
that hard or quantifiable costs can at times
loom larger than intangibles simply because
of the difficulty of valuing intangibles. 5. Does the use of the technology raise
One key in conducting this assessment is transparency-related concerns, either in
thinking through downstream or unintended how Axon communicates with members
consequences. Although the questions we of the public or how police agencies
ask will necessarily depend on the nature of engage with members of the public?
the technology, for most use cases we ask:
6. Does the use of the technology raise
1. Once deployed, can the technology risks of directly or indirectly violating
be used (or misused) in ways other constitutional or other legal rights,
than contemplated in this use case? including but not limited to: unlawful
searches, unlawful seizures/arrests,
excessive force, discovery/disclosure
2. Will this use of this technology lead violations (such as Brady, Giglio, Rosario,
to greater criminalization (people being etc.), or First Amendment concerns?
stopped, ticketed, arrested, or incarcerated)
or to policing in counterproductive ways?
7. Are there other potential social costs
that have not yet been considered,
3. How will this use of technology including but not limited to:
impact personal information • Whether there might be a unique impact
privacy? Be sure to consider: on any specific subgroup (e.g., children,
• What data are captured and from whom? LGBTQ communities, socioeconomically
How are they aggregated and/or mined? disadvantaged communities)?
• What are the data retention practices? • Whether there are historic considerations
• Who owns the data? Who has access, that may make particular communities
and what are the security practices? distrustful of this technology?
• The potential for mission creep (either over
time or in response to critical events)?
4. Does the use of the technology raise
• The impact of how others in the industry
concerns regarding racial or other identity
will respond to Axon’s precedent?
bias or disparities? Be sure to consider:
• Any global/international
• Disparities in design (e.g., whether the human rights impact?
technology itself has any inherent bias,
including algorithmic bias relating to
personal identity, for example, by employing

48
References
1
See BRAD SMITH & CAROL ANNE BROWNE, TOOLS AND WEAPONS: THE PROMISE AND THE PERIL OF THE DIGITAL AGE 211–30 (2019) (calling
for government regulation and industry self-regulation to govern growth of A.I. technologies); Amanda Askell, Miles Brundage & Gillian Hadfield,
The Role of Cooperation in Responsible AI Development, OPEN A.I. (July 10, 2019), https://arxiv.org/pdf/1907.04534.pdf (arguing that “competitive
pressures could incentivize AI companies to underinvest in ensuring their systems are safe, secure, and have a positive social impact”).
2
See DEP’T OF HOMELAND SEC., DHS EXHIBIT 300 PUB. RELEASE BY10 / CBP - LICENSE PLATE READER (LPR) 1 (2010), https://web.archive.org/
web/20110707053748/http://www.dhs.gov/xlibrary/assets/mgmt/e300-cbp-lpr2010.pdf (“The initial vision for a license plate reading tool was
developed by the U.S. Customs Service and was implemented in partnership with the Immigration and Naturalization Service through a contract
beginning in 1998.”).
3
See KEITH GIERLACK ET AL., RAND, LICENSE PLATE READERS FOR LAW ENFORCEMENT 9 (2014) https://www.ncjrs.gov/pdffiles1/nij/
grants/247283.pdf (reporting that between 83 and 91 percent of agencies use ALPRs to track auto theft).
4
See id. at 8. See also CYNTHIA LUM ET AL., THE RAPID DIFFUSION OF LICENSE PLATE READERS IN U.S. LAW ENFORCEMENT AGENCIES 10 (2016),
https://cebcp.org/wp-content/lpr/LPR-National-Survey-Report-2016.pdf (estimating 66% of agencies with 100 or more officers had ALPRs as of
Dec 2016) [hereinafter “RAPID DIFFUSION OF LPRS”]; Leonardo Expands ELSAG Automatic License Plate Recognition (ALPR) Solution with Advanced
Computer Vision Software and Cloud-Based Storage, LEONARDO (May 14, 2018), https://www.leonardocompany.com/en/press-release-detail/-/
detail/elsag-targhe (stating it provides “ALPR solutions” to nearly 4,000 law enforcement customers in over 25 countries).
5
See, e.g., DIGITAL RECOGNITION NETWORK, https://drndata.com/products/ (last visited Oct. 8, 2019).
6
See, e.g., FLOCK SAFETY, https://www.flocksafety.com/product/flock-safety-faqs (last visited Oct. 8, 2019).
7
See Automated Number Plate Recognition (ANPR) System Market Size Overview by Rising Demands, Trends and Huge Business Opportunities 2019
to 2026, NEWSSTONER (Sept. 23, 2019), https://newsstoner.com/2019/09/23/automatic-number-plate-recognition-anpr-system-market-size-
overview-by-rising-demands-trends-and-huge-bussiness-opportunities-2019-to-2026/.
8
See, e.g., Julia Angwin & Jennifer Valentino-Devries, New Tracking Frontier: Your License Plates, WALL ST. J. (Sept. 29, 2012), https://web.archive.
org/web/20130124093854/http://online.wsj.com/article/SB10000872396390443995604578004723603576296.html (noting that the U.S.
Department of Homeland Security distributed more than $50 million in federal grants to law-enforcement agencies between 2008 and 2012);
RAPID DIFFUSION OF LPRS, supra note 4, at 5 (2016) (“Federal and state funding have played a key role in LPR adoption, as 40% of agencies using
LPRs funded their initial purchases with federal grants and 21% did so with state funding.”); LPD to Buy License Plate Reader with Grant Funds,
UNION-SUN J. (Aug. 16, 2017), https://www.lockportjournal.com/news/local_news/lpd-to-buy-license-plate-reader-with-grant-funds/article_
b47afb60-48b7-5d6b-a6bd-9ed8b985d1c6.html (highlighting New York State grants for ALPR acquisition); Vigilant Solutions – LPR Grant Assistance
Program, POLICE GRANTS HELP, https://www.policegrantshelp.com/Vigilant-Solutions-LPR-Grant-Assistance/ (advertising free, customized
grant help to purchase Vigilant ALPRs) (last visited Oct. 7, 2019).
9
See, e.g., Matt Hill, The Advantages of SaaS AI-enabled Surveillance Compared to Traditional ALPRs, SECURITY SALES INTEGRATION (July 8, 2019),
https://www.securitysales.com/surveillance/advantages-saas-ai-surveillance-alprs/.
10
See Justin Rohrlich, In Just Two Years, 9,000 of These Cameras Were Installed to Spy on Your Car, QUARTZ (Feb. 5, 2019), https://qz.com/1540488/
in-just-two-years-9000-of-these-cameras-were-installed-to-spy-on-your-car/. See also Josh Kaplan, License Plate Readers Are Creeping Into
Neighborhoods Across the Country, SLATE (July 10, 2019), https://slate.com/technology/2019/07/automatic-license-plate-readers-hoa-police-
openalpr.html (reporting that one private security company alone has deployed nearly 500 OpenALPR cameras, scanning nearly 1.5 million license
plates per week).
11
See CYNTHIA LUM ET AL., LICENSE PLATE RECOGNITION TECHNOLOGY (LPR): IMPACT EVALUATION AND COMMUNITY ASSESSMENT 67–69
(2010), http://www.cebcp.org/wp-content/evidence-based-policing/LPR_FINAL.pdf (discussing connection of LPR data with a secondary data
source and tertiary data mining) [hereinafter “LPR IMPACT EVALUATION”].
12
See LPR IMPACT EVALUATION, supra note 11, at 66–71.
Dave Maass, “No Cost” License Plate Readers are Turning Texas Police into Mobile Debt Collectors and Data Miners, ELECTRONIC FRONTIER
13

FOUND. (Jan. 26, 2016), https://www.eff.org/deeplinks/2016/01/no-cost-license-plate-readers-are-turning-texas-police-mobile-debt-collectors-

and [hereinafter “Mobile Debt Collectors”].
14
See, e.g., PlatesearchTM, VIGILANT SOLUTIONS, https://www.vigilantsolutions.com/products/license-plate-recognition-lpr/ (“In addition
to access to our commercial data, agencies can choose to share with other law enforcement agencies to gain access to another 1.5B detections
nationwide.”); See also Dave Maass & Beryl Lipton, What We Learned, MUCKROCK (Nov. 15, 2018), https://www.muckrock.com/news/
archives/2018/nov/15/alpr-what-we-learned/ (reporting that, of 200 agencies surveyed, only eight declined to share their ALPR data with Vigilant).
15
VaaS International Holdings is the parent company of both Vigilant Solutions and Digital Recognition Network (DRN). DRN boasts a database of 9
billion private scans, while Vigilant maintains an additional database of at least 1.5 billion law enforcement scans. See PlatesearchTM, supra note 14;
Joseph Cox, This Company Built a Private Surveillance Network. We Tracked Someone with It, VICE: MOTHERBOARD (Sept. 17, 2019), https://www.
vice.com/en_us/article/ne879z/i-tracked-someone-with-license-plate-readers-drn.

49
16
See Reduced Prices for License Plates Readers Attracts More Buyers, HOMELAND SEC. NEWS WIRE (Jan. 24, 2012), http://www.
homelandsecuritynewswire.com/srlet20120124-reduced-prices-for-license-plates-readers-attracts-more-buyers. See also LPR IMPACT
EVALUATION, supra note 11, at 25 (reporting that the lack of outside funding available to purchase ALPRs, and cost of ALPRs were the first [43%] and
third [29.9%] most cited reasons that law enforcement agencies had not yet implemented the technology); RAPID DIFFUSION OF LPRS, supra note 4,
at Fig. 11 (indicating cost as number one hurdle keeping agencies from using ALPRs).
17
See Fleet 2, AXON, https://www.axon.com/products/fleet-2 (last visited Oct. 14, 2019).
18
See Jon Schuppe, Motorola, Known for Cellphones, Is Fast Becoming a Major Player in Government Surveillance and Artificial Intelligence,
NBC NEWS (Oct. 2, 2019), https://www.nbcnews.com/news/us-news/motorola-company-known-cellphones-fast-becoming-major-player-
government-surveillance-n1059551 (describing acquisition by Motorola Solutions of VaaS International Holdings, parent company to two industry
leaders in ALPR systems, and WatchGuard Video, an industry leader in dashcams and a producer of body worn cameras); See also Press Release,
Motorola Solutions, Motorola Solutions Acquires WatchGuard, Inc., Leader in Mobile Video for Public Safety (July 11, 2019), https://newsroom.
motorolasolutions.com/news/motorola-solutions-acquires-watchguard-inc-leader-in-mobile-video-for-public-safety.htm; Press Release,
Motorola Solutions, Motorola Solutions Acquires VaaS International Holdings, Leader in Data and Image Analytics for Vehicle Location (Jan. 7,
2019), https://newsroom.motorolasolutions.com/news/motorola-solutions-acquires-vaas-international-holdings-leader-in-data-and-image-
analytics-for-vehicle-location.htm.
19
See, e.g., OHIO STATE HIGHWAY PATROL PLANNING SERVICES SECTION RESEARCH & DEVELOPMENT, AUTOMATIC PLATE READER TECHNOLOGY
11 (2005) (finding that the use of ALPRs significantly increased stolen vehicle recoveries and arrests compared to the previous year); IRWIN M. COHEN
ET AL., A REPORT ON THE UTILITY OF THE AUTOMATED LICENSE PLATE RECOGNITION SYSTEM IN BRITISH COLUMBIA 15–17 (2007) (finding greater
number of hits for uninsured, prohibited, unlicensed, or stolen vehicle drivers with use of LPRs); PA CONSULTING GROUP, DRIVING CRIME DOWN
125–30 (2004) (finding that ALPRs significantly enhanced the ability of officers to make arrests, particularly when officers were dedicated specifically
to an ALPR unit).
20
See, e.g., Jason Potts, Research in Brief: Assessing the Effectiveness of Automatic License Plate Readers, POLICE CHIEF 14–15 (Mar. 2018), https://
www.theiacp.org/sites/default/files/2018-08/March%202018%20RIB.pdf (finding that, in a trial, cars equipped with ALPRs “showed a 140%
greater ability to detect stolen cars” than cars without ALPRs).
21
See Cynthia Lum et al., Understanding the Limits of Technology’s Impact on Police Effectiveness, 20 POLICE Q. 135, 136 (2017). See generally, BRUCE
TAYLOR ET AL., COMBATTING AUTO THEFT IN ARIZONA (2011) (detailing randomized experiment in Mesa, Arizona that found a small auto theft
squad was much more likely to detect and recover stolen vehicles and make arrests when using ALPRs).
22
See ACLU, YOU ARE BEING TRACKED: HOW LICENSE PLATE READERS ARE BEING USED TO RECORD AMERICANS’ MOVEMENTS 13–15 (July
2013), https://www.aclu.org/files/assets/071613-aclu-alprreport-opt-v05.pdf (“Of the 1,691,031 plates scanned by the Minnesota State Patrol from
2009–2011, just 852 citations were issued and 131 arrests were made. That is 0.05 percent of plate reads.” (internal citations omitted)) [hereinafter
“YOU ARE BEING TRACKED”].
23
See, e.g., 7 Cases Solved Thanks to ALPR Data, POLICEONE (June 12, 2018), https://www.policeone.com/police-products/traffic-enforcement/
license-plate-readers/articles/7-cases-solved-thanks-to-alpr-data-doayALt3VGwqCIN5/ (reporting several instances where queries of historical
ALPR data helped to solve serious crimes).
24
See DAVID J. ROBERTS & MEGHANN CASANOVA, AUTOMATED LICENSE PLATE RECOGNITION SYSTEMS 22–23 (Aug. 2012), https://www.ncjrs.
gov/pdffiles1/nij/grants/239604.pdf (finding that although only 25% of agencies initially acquired ALPRs for general investigations, 70% reported
doing so in practice).
25
See Automated License Plate Readers: State Statutes, NAT’L CONF. ST. LEGISLATURES (Mar. 15, 2019), http://www.ncsl.org/research/
telecommunications-and-information-technology/state-statutes-regulating-the-use-of-automated-license-plate-readers-alpr-or-alpr-data.
aspx. See also infra Part IV.
26
See, e.g., TAYLOR ET AL., supra note 21, at 34–40; LPR IMPACT EVALUATION, supra note 11, at 5–7.
27
See, e.g., HANI S. MAHMASSANI ET AL., CHICAGO RED LIGHT CAMERA ENFORCEMENT 43–44 (Mar. 2017) (showing red light cameras significantly
reduce fatal crashes and that this effect is persistent over time); FEDERAL HIGHWAY ADMINISTRATION, U.S. DEP’T OF TRANSPORT., SAFETY
EVALUATION OF RED-LIGHT CAMERAS 5–6 (Apr. 2005) (finding small but significant overall benefit of red light cameras, concluding that because
rear-end crashes are generally less severe and lower cost, they do not offset the positive effects of the cameras).
28
These findings come from a single study of small towns in Massachusetts, and it is unclear if these results would hold under different
circumstances. See Michael D. Makowsky & Thomas Stratmann. More Tickets, Fewer Accidents: How Cash-Strapped Towns Make for Safer Roads, 54
J.L. & ECON. 863–88 (2011).

50
29
The literature and research on this subject are vast. See Whren v. United States, 517 U.S. 806 (1996) (declaring that any traffic offense committed by
a driver was a legitimate basis for a stop, even if that offense was pretextual); Heien v. North Carolina, 135 S. Ct. 530 (2014) (holding that an officer’s
reasonable mistake of the law can provide the individualized suspicion necessary for a legitimate traffic stop); Ken Armstrong, How the Supreme
Court Made it Legal for Cops to Pull You Over for Just About Anything, THE MARSHALL PROJECT (Aug. 3, 2015), https://www.themarshallproject.
org/2015/08/03/how-the-supreme-court-made-it-legal-for-cops-to-pull-you-over-for-just-about-anything (reporting on the expanded
discretion police can exercise in performing pretextual stops, even when the pretext is not a legal justification for a stop); See also, e.g., FRANK R.
BAUMGARTNER, SUSPECT CITIZENS 77 (2018) (finding—based on the study of 20 million traffic stops in North Carolina—“strong, consistent, and
powerful evidence that black and white drivers face dramatically different odds of being pulled over,” and noting that, in conservative estimate, black
drivers were 63% more likely to be pulled over than whites); EMMA PIERSON, ET AL., A LARGE-SCALE ANALYSIS OF RACIAL DISPARITIES IN POLICE
STOPS ACROSS THE UNITED STATES 3–5 (2019), https://5harad.com/papers/100M-stops.pdf (finding—based on the study of nearly 100 million
stops from 21 state patrol agencies and 29 municipal police departments—that black drivers were more likely to be stopped during daylight hours,
when their race was more readily apparent, than at night, and that in all cases they experienced higher rates of stops than white drivers under similar
circumstances).
30
See NAT’L HIGHWAY TRAFFIC SAFETY ADMIN., Characteristics of Law Enforcement Officers’ Fatalities in Motor Vehicle Crashes 22 (2011) (finding
that, from 1982–2008, “inattentive/careless driving [talking, eating, car phone, etc.]” accounted for 29 law enforcement officer crash fatalities, while
“failure to keep in proper lane or running off the road” accounted for 225 fatalities, speeding accounted for 220 fatalities, and high-speed chases
resulted in 69 fatalities).
31
See Chief Joel F. Shults, How ALPR Data Drives Intelligence and Policing, POLICE ONE (May 3, 2018), https://www.policeone.com/police-
products/traffic-enforcement/license-plate-readers/articles/473999006-How-ALPR-data-drives-intelligence-led-policing/ (“ALPR results,
triangulated with other statistical data, can reveal patterns of activity associated with criminal events that analysts can use to establish probabilities
of crimes and their locations.”); See also Elizabeth E. Joh, The New Surveillance Discretion: Automated Suspicion, Big Data, and Policing, 10 HARV. L. &
POL. REV. 15, 23 (2016) (noting that ALPR geo-fencing may be used to track “suspicious” vehicle patterns and behaviors).
32
See, e.g., ROBERTS & CASANOVA, supra note 24, at 23 (highlighting a case study in which a single officer equipped with an ALPR device for only
96 hours over the course of a month was able to read 48,101 plates, leading to 255 traffic citations, the identification of 26 drivers with suspended
licenses, 16 vehicle emission violators, 4 stolen and 1 expired license plate, and 3 arrests).

See, e.g., PRESIDENT’S TASK FORCE ON 21ST CENTURY POLICING, FINAL REPORT OF THE PRESIDENT’S TASK FORCE ON 21ST CENTURY
33

POLICING 9–18 (2015), https://cops.usdoj.gov/RIC/Publications/cops-p311-pub.pdf (“Pillar 1: Building Trust & Legitimacy”).

See GIERLACK ET AL, supra note 3, at 50 (explaining that for some agencies that use ALPRs for traffic enforcement, alerts were so frequent—as
34

much as every few seconds—that officers were forced either to ignore the alerts or turn off alerts for expired registrations and suspended licenses).
35
Dave Maass & Jeremy Gillula, ELECTRONIC FRONTIER FOUND, What You Can Learn From Oakland’s Raw ALPR Data (Jan. 21, 2015), https://www.
eff.org/deeplinks/2015/01/what-we-learned-oakland-raw-alpr-data (finding ALPR scans were also far more likely to occur in lower-income
neighborhoods).
36
See, e.g., MARIO SALAS & ANGELA CIOLFI, DRIVEN BY DOLLARS: A STATE-BY-STATE ANALYSIS OF DRIVER’S LICENSE SUSPENSION LAWS FOR
FAILURE TO PAY COURT DEBT 2, 14–15 (2017) (finding that 43 states suspend driver’s licenses for failure to pay a court debt); See also Brandon
Garrett, When the Police Come for Your Driver’s License, AM. CONSERVATIVE (Apr. 9, 2019), https://tinyurl.com/y5qr6bpl (finding nearly one in
seven adult drivers in North Carolina has had their licenses suspended for a reason unrelated to driving).
37
See, e.g., Samar Khurshid, NYPD Continues to Move Away from Criminal Penalties for Low-Level Offenses, but Racial Disparities Remain, GOTHAM
GAZETTE (Sept. 4, 2019), https://www.gothamgazette.com/city/8768-nypd-fewer-criminal-penalties-for-low-level-offenses-racial-differences-
remain (reporting that—following the adoption of new guidelines—although the NYPD issued 90% fewer criminal summonses for low-level
offenses over the previous year, 91% of those issued were to people of color).
38
See, e.g., William Farell, Predominately Black Neighborhoods in D.C. Bear the Brunt of Automated Traffic Enforcement, D.C. POL’Y CTR. (June
28, 2018), https://www.dcpolicycenter.org/publications/predominately-black-neighborhoods-in-d-c-bear-the-brunt-of-automated-traffic-
enforcement/ (finding that there were more speed and red light cameras placed—and therefore more citations issued—in predominantly Black
neighborhoods than in White or mixed-race neighborhoods, despite equal incidence of car crashes).
39
See, e.g., YOU ARE BEING TRACKED, supra note 22, at 13–15.
40
See, e.g., Automatic License Plate Recognition (ALPR), MOTOROLA SOLUTIONS, https://www.creativecom.com/public-safety/alpr.htm (last
accessed Sep. 25, 2019).
41
See, e.g., U.S. DEP’T OF JUST., INVESTIGATION OF THE FERGUSON POLICE DEPARTMENT 4–5 (Mar. 2015) (summarizing findings of racial
disparities, including several measures relating to fines and fees enforcement) [hereinafter “FERGUSON REPORT”]; Kathryn Zickuhr, Applying a Racial
Equity Lens to Fines and Fees in the District of Columbia, D.C. POL’Y CTR. (Apr. 22, 2019), https://www.dcpolicycenter.org/publications/racial-
equity-fines-fees/ (finding that unpaid fines from traffic tickets in Washington D.C. are doubled after 30 days; are sent to collections after 90 days,
where a 20% surcharge is added; and that ultimately low-income drivers “may be forced to choose between losing their source of income and driving
with an expired license—increasing the likelihood that they could face additional debt or even imprisonment”).
42
Mike Maciag, Addicted to Fines: Small Towns in Much of the Country are Dangerously Dependent on Punitive Fines and Fees, GOVERNING (Sept.
2019), https://www.governing.com/topics/finance/gov-addicted-to-fines.html.

51
43
See S.B. 5, 98th Gen. Assemb., Reg. Sess., at 5 (Mo. 2015), available at https://www.senate.mo.gov/15info/pdf-bill/tat/SB5.pdf; FERGUSON
REPORT, supra note 41, at 2. The Ferguson Police Department practice of citing residents with as many municipal infractions as possible to incur the
maximum amount of fines and fees has since become the prototypical example of “fines and fees” policing. See generally Jessica Brand, How Fines
and Fees Criminalize Poverty: Explained, APPEAL (July 16, 2018), https://theappeal.org/fines-and-fees-explained-bf4e05d188bf/.
44
See Mobile Debt Collectors, supra note 13; see also, e.g., George Joseph, What Are License-Plate Readers Good For?, CITY LAB (Aug. 5, 2016),
https://www.citylab.com/equity/2016/08/what-are-license-plate-readers-good-for/492083/ (reporting that ALPRs have been used for fines
and fees policing in at least three Texas communities); Eric Markowitz, Pay This Fee, Or Go to Jail, INT’L BUS. TIMES (Feb. 3, 2016), https://www.
ibtimes.com/pay-fee-or-go-jail-how-license-plate-scanner-vigilant-solutions-makes-money-texas-2290835 (reporting that New York City
has used ALPRs to send nearly 35,000 parking summonses in 2011, and that in the first 12 hours of use of ALPRs in New Haven, Connecticut police
identified 119 vehicles with unpaid parking violations, generating $40,000 in city fines paid); cf. Office of the Assistant Sec’y Research & Tech.,
Dep’t of Transp., In Arizona, an Automatic License Plate Recognition (ALPR) System Has an Estimated Benefit-to-Cost Ratio of 9.6 Due to Improved
Vehicle Registration and Insurance Compliance, INTELLIGENT TRANS. SYSTEMS PROG. OFF. (June 2008), https://www.itsbenefits.its.dot.gov/ITS/
benecost.nsf/ID/20746487B947B3358525797C006528CF?OpenDocument&Query=BMeasure (suggesting that the implementation of ALPRs for
use in registration enforcement in Arizona could generate “15,557,500 recovered in registration fees and $18,459,907 in fines”).
45
See Potts, supra note 20, at 14–15 (finding that, in a trial, 35% of all mobile ALPR “hits” were misreads, with 37% for fixed readers).
46
See, e.g., Cyrus Farivar, Due to License Plate Reader Error, Cop Approaches Innocent Man, Weapon in Hand, ARS TECHNICA (Apr. 23, 2014), https://
arstechnica.com/tech-policy/2014/04/due-to-license-plate-reader-error-cop-approaches-innocent-man-weapon-in-hand/ (detailing an instance
in which an ALPR misread a “7” for a “2” and improperly alerted the officer that the vehicle was stolen, leading the officer to approach the vehicle
with a drawn gun); Kade Crockford, San Francisco Woman Pulled Out of Car at Gunpoint Because of License Plate Reader Error, ACLU (May 13, 2014),
https://www.aclu.org/blog/privacy-technology/location-tracking/san-francisco-woman-pulled-out-car-gunpoint-because (citing example of
San Francisco woman who was made to exit her car, kneel, and was handcuffed as multiple officers held her at gunpoint—one with a shotgun—
when an ALPR misread her plate and listed her car as stolen).
47
See, e.g., Lisa Fernandez, Privacy Advocate Sues CoCo Sheriff’s Deputies After License Plate Readers Target His Car Stolen (Feb. 19, 2019), https://
www.ktvu.com/news/privacy-advocate-sues-coco-sheriffs-deputies-after-license-plate-readers-target-his-car-stolen (reporting that an Oakland
man and his brother were detained at gunpoint when an ALPR reported their rental car as stolen.—although the car had been recovered by the rental
company four months prior, the ALPR provider’s database had not been updated).
48
For example, in both Herring v. United States, and Arizona v. Evans, the police searched the defendants’ vehicles, and arrested them for
contraband, on the basis of erroneous warrants which resulted from clerical errors. See Herring v. United States, 555 U.S. 135, 155 (2009) (Ginsburg,
J., dissenting) (noting that “[e]lectronic databases form the nervous system of contemporary criminal justice,” and that “inaccuracies in expansive,
interconnected collections of electronic information raise grave concerns for individual liberty.”); Arizona v. Evans, 514 U.S. 1 (1995) (Stevens, J.,
dissenting) (“The offense to the dignity of the citizen who is arrested, handcuffed, and searched on a public street simply because some bureaucrat
has failed to maintain an accurate computer data base strikes me as . . . outrageous.”); see also, e.g., Jack Leonard, ID Errors Put Hundreds in County
Jail, L.A. TIMES (Dec. 25, 2011), https://www.latimes.com/archives/la-xpm-2011-dec-25-la-me-wrong-id-20111225-story.html (reporting that in
Los Angeles, hundreds of people were wrongly imprisoned due either to errors in warrant databases, or mistakes made by police while executing a
warrant); Wayne J. Pitts, From the Benches and Trenches: Dealing With Outstanding Warrants for Deceased Individuals, 30 JUST. SYS. J. 219 (2009) (“A
number of data-reliability problems in the warrant database soon became apparent as there were examples of illogical birth dates. Similarly, Social
Security numbers were often inverted, missing, incomplete, or otherwise inaccurate. There were also problems with misspellings, different spellings,
or both of names listed in the warrant database.”).
49
See, e.g., Home Page, TURO, https://turo.com/en-us (“Discover the world’s largest car sharing marketplace.”) (last visited Oct. 9, 2019); Ankita
Bhutani & Pallavi Bhardwaj, Carsharing Market Size by Model, Growth Potential, Competitive Market Share & Forecast 2018–2024, GLOBAL
MARKETING INSIGHTS, https://www.gminsights.com/industry-analysis/carsharing-market (forecasting that the car-sharing market is expected to
grow to $12 billion by 2024) (last updated July 2019).
50
See Cox, supra note 15.
51
Id.
52
“[T]he retrospective quality of the data here gives police access to a category of information otherwise unknowable. In the past, attempts to
reconstruct a person’s movements were limited by a dearth of records and the frailties of recollection. With access to [cell-site location information],
the Government can now travel back in time to retrace a person’s whereabouts . . . . Critically. . . this newfound tracking capacity runs against
everyone. Unlike with the GPS device in Jones, police need not even know in advance whether they want to follow a particular individual, or
when.” Carpenter v. United States, 138 S. Ct. 2206, 2218 (2018); see also United States v. Jones, 565 U.S. 400, 415 (2012) (Sotomayor, J., concurring)
(“[Location] monitoring generates a precise, comprehensive record of a person’s public movements that reflects a wealth of detail about her familial,
political, professional, religious, and sexual associations.”).
53
See Carpenter v. United States, 138 S. Ct. 2206, 2221 (2018) (holding that the government must generally obtain a warrant prior to searching
historical cell-site location information records).

52
54
See, e.g., Adam Goldman & Matt Apuzzo, With Cameras, Informants, NYPD Eyed Mosques, ASSOCIATED PRESS (Feb. 23, 2012), https://www.
ap.org/ap-in-the-news/2012/with-cameras-informants-nypd-eyed-mosques (discussing NYPD tracking all of the vehicles outside of particular
mosques); Brian M. Rosenthal, Anti-Abortion Activists Adopt a New Tactic: Tracking License Plates, HOUS. CHRON. (Aug. 13, 2014), https://www.
houstonchronicle.com/news/politics/texas/article/Anti-abortion-activists-adopt-a-new-tactic-5687420.php (reporting on the safety risks and
privacy risks experienced by people entering abortion clinics whose license plates are tracked by anti-abortion activists). See generally Linda M.
Merola & Cynthia Lum, Predicting Public Support for the Use of License Plate Recognition Technology by Police, 15 POLICE PRAC. & RES. 373, 375
(2014) (noting concern that “widespread LPR use might lead individuals to suppress unpopular, unconventional, or embarrassing actions even if these
acts are not illegal”).
55
See Sadie Gurman, AP: Across US, Police Officers Abuse Confidential Databases, ASSOCIATED PRESS (Sept. 28, 2016), https://apnews.
com/699236946e3140659fff8a2362e16f43; Michael Powell et al., Lt. Stowe’s Sudden Fall From Grace, WASH. POST (Nov. 30, 1997), https://www.
washingtonpost.com/archive/politics/1997/11/30/lt-stowes-sudden-fall-from-grace/a6ac37f2-57d2-47fb-b6da-0f8f6a45dde8/ (reporting on
officer who pleaded guilty to using an ALPR database to blackmail married men who parked outside a gay club).
56
Cooper Quintin & Dave Maass, License Plate Readers Exposed! How Public Safety Agencies Responded to Major Vulnerabilities in Vehicle
Surveillance Tech, ELECTRONIC FRONTIER FOUND. (Oct. 28, 2015), https://www.eff.org/deeplinks/2015/10/license-plate-readers-exposed-
how-public-safety-agencies-responded-massive; see also Kenneth Lipp, License to Connive: Boston Still Track Vehicles, Lies About It, and Leaves
Sensitive Resident Data Exposed Online, DIG BOS. (Sept. 8, 2015), https://digboston.com/license-to-connive-boston-still-tracks-vehicles-lies-
about-it-and-leaves-sensitive-resident-data-exposed-online/ (reporting that the City of Boston was hosting nearly 1 million license plate scans
dating back three years in a freely-accessible server).
57
See Zack Whittaker, Police License Plate Readers Are Still Exposed on the Internet, TECH CRUNCH (Jan. 22, 2019), https://techcrunch.
com/2019/01/22/police-alpr-license-plate-readers-accessible-internet/.
58
See, e.g., RAPID DIFFUSION OF LPRS, supra note 4, at Fig. 8 (indicating that over 30% of agencies surveyed had LPRs regularly access information
about “vehicles related to documented gang members”); Mark Harris, If You Drive in Los Angeles, the Cops Can Track Your Every Move, WIRED (Nov.
13, 2018), https://www.wired.com/story/drive-los-angeles-police-track-every-move/ (reporting that the LAPD cross-referenced historical ALPR
data against a gang database to obtain a search warrant connected to a murder).
59
See e.g., CITY OF CHICAGO OFFICE OF INSPECTOR GENERAL, REVIEW OF CHICAGO POLICE DEPARTMENT’S “GANG DATABASE” 2 (Apr.
2019), https://igchicago.org/wp-content/uploads/2019/04/OIG-CPD-Gang-Database-Review.pdf (concluding that “CPD’s gang information
contains incomplete and contradictory data;” “CPD does not regularly review, correct, or purge inaccurate gang information; those with inaccurate
designations have no opportunity to clear their name and mitigate the impact of incorrect or outdated gang designations;” and “CPD’s gang
designations are permanent and inescapable”).
60
See National Crime Information Center (NCIC) — NCIC Files, https://www.fbi.gov/services/cjis/ncic (showing that Immigration Violator File
includes “[r]ecords on criminal aliens whom immigration authorities have deported and aliens with outstanding administrative warrants of removal”).

See, e.g., County of Sacramento, Change to Delegated Purchase Order DP81191041, Vigilant Contract 9 (June 2016), https://www.documentcloud.
61

org/documents/4618380-ITEM-4-Contract-No-DP81191041-FY-16-17.html#document/p9/a443654.
62
See Automatic Number Plate Recognition, POLICE.UK, https://www.police.uk/information-and-advice/automatic-number-plate-recognition/
(providing information on why police use ALPR, how it works, retention and storage policies, camera locations, and links to further information)
(last visited Oct. 9, 2019). Many of the protections were put in place to comply with the European Convention on Human Rights, and supporting U.K.
statutes such as the Data Protection Act 1998 (DPA) and the Regulation of Investigatory Powers Act 2000 (RIPA). See HOME OFFICE, THE USE OF
ANPR BY LAW ENFORCEMENT AGENCIES 2 (2014), https://www.npcc.police.uk/documents/The%20Use%20of%20ANPR%20by%20Law%20
Enforcement%20Agencies.pdf.
63
See Julia M. Brooks, Drawing the Lines: Regulation of Automatic License Plate Readers in Virginia, 25 RICH. J.L. & TECH. 1, 19 (2019) (finding that
only two bills addressing ALPRs have been introduced to Congress—both of which limited data retention to 30 days—though they perished in
committee).
64
18 U.S.C. § 2721 et seq; see also Dave Maass, Mystery Show Debunks License Plate Privacy “Myth”, ELECTRONIC FRONTIER FOUND. (June 15, 2015),
https://www.eff.org/deeplinks/2015/06/mystery-show-podcast-debunks-license-plate-privacy-myth (critiquing the Drivers Privacy Protection
Act for its many exceptions and lack of meaningful protections).
65
See Automated License Plate Readers: State Statutes, NAT’L CONF. ST. LEGISLATURES (Mar. 15, 2019), http://www.ncsl.org/research/
telecommunications-and-information-technology/state-statutes-regulating-the-use-of-automated-license-plate-readers-alpr-or-alpr-data.
aspx (listing 16 states with ALPR statutes). See also STATE OF FLA., CRIMINAL AND JUVENILE JUSTICE INFO. SYS. COUNCIL, GUIDELINES FOR THE
USE OF AUTOMATED LICENSE PLATE READERS, https://www.fdle.state.fl.us/CJJIS/Documents/CJJIS-Council-ALPR-Guidelines (last visited Oct.
9, 2019) [hereinafter “FLORIDA CJJIS GUIDELINES”]; N.Y. STATE DIV. CRIMINAL JUSTICE SERVS., SUGGESTED GUIDELINES: OPERATION OF LICENSE
PLATE READY TECHNOLOGY (2011), https://apps.criminaljustice.ny.gov/ofpa/pdfdocs/finallprguidelines01272011a.pdf.
66
N.H. REV. STAT. ANN. § 261.75-b (VIII).
67
ME. REV. STAT. ANN. § 2117-A(2) (21 days); MINN STAT. ANN. § 13.824(3)(a) (60 days); CAL. VEH. CODE § 2413(b) (restricting California Highway Patrol
scans to 60 days retention); N.C. GEN. STAT. § 20-183.32 (90 days); MONT. CODE ANN. § 46-5-118(1) (90 days); TENN. CODE § 55-10-302 (90 days).
68
COLO. REV. STAT. § 24-72-113 (2014) (3 years); FLORIDA CJJIS GUIDELINES, supra note 65, at 6(e) (3 years); GA. CODE § 35-1-22 (30 months); VT.
STAT. ANN. tit. 23, § 1607(d)(2) (18 months). Agency policies similarly vary widely. See RAPID DIFFUSION OF LPRS, supra note 4, at Fig. 10 (showing
wide range of retention periods in agency policies, with 17.6% indicating 1 month or less, and 15% indicating indefinite retention).

53
69
Neal v. Fairfax Cty. Police Dep’t, 295 Va. 334, 350 (Va. 2018).
70
See, e.g., UTAH CODE ANN. § 41-6a-2003; N.C. GEN. STAT. § 20-183.31; VT. STAT. ANN. tit. 23, § 1607(a)(5). The better (but rarer) approach is to
enumerate specific permissible uses. See, e.g., MONT. CODE ANN. § 46-5-117(2)(d)(v); NEB. REV. STAT. 60-3203(2); N.H. REV. STAT. ANN. §§ 261.75-
b(V). But even these statutes do not strictly curtail enforcement of low-level offenses or limit enforcement overall.
71
See, e.g., GA. CODE § 35-1-22; N.C. GEN. STAT. § 20-183.31; MONT. CODE ANN. § 46-5-117(2)(d)(i); NEB. REV. STAT. § 60-3206(1).
CAL. CIV. CODE § 1798.29; see also NEB. REV. STAT. § 60-3206(2) (requiring agencies adopt “a privacy policy to ensure that captured plate data is
72

not shared in violation of the Automatic License Plate Reader Privacy Act or any other law and conspicuously post the privacy policy on its Internet
web site”).
73
See Maass & Gillula, supra note 35.
74
See, e.g., Kim Zetter, Police Contract With Spy Tool Maker Prohibits Talking About Device’s Use, WIRED (Mar. 4, 2014), https://www.wired.
com/2014/03/harris-stingray-nda/ (reporting on required non-disclosure agreement in Harris Corporation contracts that barred police agencies
from discussing their use of their cell-site simulators); Michael Price & Emily Hockett, Palantir Contract Dispute Exposes NYPD’s Lack of Transparency,
Brennan Center for Justice (Jul. 20, 2017), https://www.brennancenter.org/our-work/analysis-opinion/palantir-contract-dispute-exposes-
nypds-lack-transparency (reporting that Palantir’s contract with the NYPD allowed Palantir to “retain all rights” to data collected by the NYPD
for law enforcement purposes); Law Enforcement Archival Reporting Network – LEARN Hosted Database Use Agreement, VIGILANT SOLUTIONS
(2014), https://www.wired.com/wp-content/uploads/2014/04/LEARN_Hosted_Server_User_Agreement.pdf (prohibiting law enforcement from
disclosing the contents of their contract or “to voluntarily provide ANY information, including interviews, related to LEARN products or its services to
any member of the media” without express written consent from Vigilant).
75
See Printz v. United States, 521 U.S. 989 (1997) (holding that Congress cannot commandeer state officers to enforce federal regulations); see
also METROPOLITAN POLICE DEPARTMENT, EO-17-010, Administrative Warrants in NCIC (eff. March 24 2017), https://go.mpdconline.com/GO/
EO_17_010.pdf (directing officers to ignore NCIC alerts solely for administrative removal warrants); But see TEX. S.B. 4 § 752.053 (Texas statute
barring local law enforcement agencies or campus police forces from adopting, enforcing, or endorsing a policy which “prohibits or materially limits
the enforcement of immigration laws.”).
76
The federal NCIC, for example, “requires that records entered into the vehicles hot list be double checked by a second party to verify that the
data entered matches that contained in an investigative file.” INT’L ASS’N OF POLICE CHIEFS, PRIVACY IMPACT ASSESSMENT REPORT FOR
THE UTILIZATION OF LICENSE PLATE READERS 46 (Sept. 2009), https://www.theiacp.org/sites/default/files/all/k-m/LPR_Privacy_Impact_
Assessment.pdf.
77
The Supreme Court may soon address this issue at a matter of constitutional law. See Kansas v. Glover, Docket No. 18-556 (U.S.) (oral argument
scheduled for November 4, 2019).

See, e.g., HOME OFFICE, NATIONAL ANPR STANDARDS FOR POLICING AND LAW ENFORCEMENT ANPR 28–31 (Jan. 2019), https://assets.
78

publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/806674/NASPLE_-_January_2019_.pdf (“9.3
Organisational and User Access to ANPR Data”); HOME OFFICE, NATIONAL STANDARDS FOR COMPLIANCE AND AUDIT OF LAW ENFORCEMENT
ANPR 6–9 (Jul. 2019), https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/814216/ANPR_
Compliance_and_Audit_Standards_V1.1_July_2019.pdf (“8 Data Access and Management Standards Audit”).
79
For the most part, federal courts have had little to say about ALPR use. Because license plates are in public view, federal law currently offers
no protection under the Fourth Amendment, though there are some signs this may eventually change. In Carpenter v. United States, 138 S. Ct.
2206 (2018), the Supreme Court held that warrants are required for law enforcement inquiries about mobile phone location over seven days. This
decision may bear upon the use of ALPRs for historical investigations. Indeed, there is a portion of the Board that believes that Carpenter’s warrant
requirement should be applied to ALPRs, and there is some other litigation in the offing. See, e.g., Brief for Electronic Frontier Foundation et al. as
Amici Curiae Supporting Appellant, United States v. Yang, No. 18-10341 (9th Cir. 2019). Although not yet a settled legal question, building this into the
design would be a productive step.
80
See supra notes 56–57.
81
See Brooks, supra note 63, at 10–11 (“Billions of scans collected by private individuals, without any concern about the constitutionality of their
actions, are accessible to the thousands of participating law enforcement agencies.”); Maass & Lipton, supra note 14 (“It’s unclear which policies and
laws govern ALPR data when it crosses state boundaries. For example, a sheriff’s office in California may require officers to attach a case number to
every search of ALPR data, whereas a police department in Georgia may not have similar requirements. Meanwhile, Georgia law requires ALPR data
to be destroyed after 30 months, whereas other states may allow agencies to hold onto the data indefinitely.”).
82
Our First Report detailed how the Board operates. Our membership has evolved over time and our current members are listed on Axon’s website:
https://www.axon.com/info/ai-ethics. In keeping with the disclosures made in our First Report, we note that the Policing Project at New York
University School of Law, which staffs the Board, received $30,000 from Axon in order to defray some of its expenses in staffing the Board’s ALPR
work and preparing this report. More information on the Policing Project is available at www.policingproject.org.

54