We Tracked the Secret Police Microphones Hidden Everywhere

These controversial surveillance devices are designed

to alert authorities to gunshots,

and their exact locations have been kept hidden

from police and the public until now.

WIRED obtained leaked documents

that reveal for the first time the secret locations

of 25,580 ShotSpotter microphones.

In this video, we'll analyze that data

and test the claim of activists that these sensors lead

to biased over-policing

of communities of color across America.

12 million Americans live in a neighborhood

with at least one ShotSpotter microphone.

Are you one of them?

Let's put ShotSpotter secret locations on the grid.

[mellow music]

[graphics whirring] [mellow music]

This is just one of the 25,580 data points

that represent ShotSpotter microphone locations

as indicated in the leaked documents

provided to me earlier this year

by a source under the condition of anonymity.

To confirm that the leaked data was legit,

WIRED vetted the locations

by physically visiting sensors in different cities,

including Pasadena, California, Chicago, Illinois.

[Reporter] I think that's it on top of the pole.

[Dhruv] Miami, Florida.

[Reporter] There it is. I see it.

And this one attached to a street lamp near Prospect Park

in Brooklyn, New York,

where we spotted the protective casing

that houses the acoustic sensors and processors.

We even used Google Street View

to virtually visit a random sample

of locations in the document,

and they check out.

The sensors were exactly where the leaked data

said that they would be.

We don't know if our data set

includes all sensors that exist,

but Tom Chittum from SoundThinking,

the company that makes them,

said that as of February 2023,

the document was likely authentic.

The data set represents

over a thousand elementary and high schools,

dozens of billboards, scores of hospitals,

and more than a hundred public housing complexes

where the sensors are placed.

In total, the leaked document

indicates ShotSpotter locations

in 84 metropolitan areas across 34 states,

plus US territories such as Puerto Rico

and the Virgin Islands.

Nine US cities actually have

more than 500 sensors installed,

including Albuquerque, New Mexico,

Chicago, Illinois, Las Vegas, Nevada,

and Washington, D.C.,

where sensors can be found on US government buildings,

including the headquarters of the FBI,

the Department of Justice, the US Court of Appeals,

and just outside the city

throughout the campus of the University of Maryland.

So why are specific locations chosen for ShotSpotter?

According to Tom Chittum,

when police departments purchase SoundThinking services,

they provide the company with data

about gun violence in the area.

Once a plan is agreed upon by the department,

SoundThinking will seek permission

from private property owners, utility companies,

and business owners

to install sensors on their premises.

While Chittum says that most property owners

agree to do so without needing incentivization,

the company will occasionally offer gift cards

to secure access to their private property.

According to the company,

if a loud, impulsive sound is heard in the coverage area,

sensors often on top of buildings

and streetlights pick it up.

The location of the incident is determined

by measuring the time the sound takes

to travel to each sensor.

At minimum, three sensors need to detect a sound

for it to be flagged.

The sound is then first analyzed by AI

to determine if it's likely a gunshot

before being sent to a ShotSpotter incident review center

where analysts determine whether it's gunfire

or something else,

like a firework or a car backfiring.

If it's a gunshot,

ShotSpotter then alerts law enforcement.

SoundThinking claims they keep the exact locations

of their sensors secret, even from police,

to protect people who allow the sensors on their property

as well as to prevent the mics themselves

from being tampered with or vandalized.

But this secrecy has been a point of contention

for those who criticize the company.

[Protestors] DPD we want you out.

[Dhruv] The ACLU argues that, quote,

There are deep problems

with the gunshot detection company

and its technology,

including its methodology, effectiveness,

and impact on communities of color.

So are activists right?

Where are these microphones being placed?

WIRED did some digging and found the answer.

After receiving the leaked documents

of the sensor locations,

we used the most recent census data

and collected the demographic information

from every census block group

with at least one SoundThinking sensor.

Each census block group

has between 600 and 3,000 people in it.

We then analyzed sensor distribution in US cities

and found that, in aggregate, 70% of people

who live in a neighborhood

with at least one SoundThinking sensor

identified as either Black or Latino,

nearly 3/4 of these neighborhoods are majority non-white,

and the average household earns

a little more than 50,000 a year.

But let's zoom in and look at individual jurisdictions.

The pattern appears to be the same.

Winston-Salem, North Carolina, for example,

is approximately 43% white,

but only 13% of residents who live in areas

that have at least one ShotSpotter sensor

identified as white.

In Fort Myers, Florida,

roughly 19% of the population is Black.

In block groups with at least one sensor, however,

approximately 41% of the population is Black.

So WIRED's findings do align

with the theory put forth by activists and critics

that ShotSpotter surveillance does disproportionately occur

in poor communities of color.

Why invest in ShotSpotter

when it's shown it doesn't curb gun violence,

it doesn't do what it says it's supposed to do?

So what happens when police answer a call

based on a ShotSpotter report in one of these neighborhoods?

Let's go to Cincinnati, Ohio.

It's New Year's Eve 2022, 8:21 p.m.

Several sensors pick up two loud sounds

that could be gunshots.

This is body cam footage

that WIRED obtained of the incident.

You can see police detaining a man

who happened to be standing near the corner

where the ShotSpotter mics detected supposed gunfire.

Nine officers were sent to the location to respond.

However, once there, they found no gun,

no bullet casings, no bullet holes,

nor any evidence that a crime had been committed there.

[Officer] So we kind of did about 10 feet out

and there's nothing, unless he had a revolver.

Yet cops still arrested man after they ran his name

and found he'd failed to appear in court

for traffic violations.

This incident tracks with a 2021

Northwestern University School of Law study

which found that over a two-year period,

89% of alerts in the city of Chicago did not result

in police finding evidence of a gun-related crime.

This is likely at least in part due

to how often suspected gunfire alerts

end up being something else entirely.

For example, according to this article,

officers in Pasadena responding

to scenes of ShotSpotter alerts

have pointed to backfiring cars, fireworks,

or even noisy construction as the actual sources

of the sounds the sensors detected,

and that's when the sensors were actually working.

[mellow music]

In December 2022, SoundThinking sensors failed

to issue an alert for a shooting at a gyros shop in Chicago

that wounded two men.

Reportedly, 55 rounds were fired,

yet SoundThinking's equipment sent out no alerts.

The missed detections were apparently due

to three out-of-service sensors.

In fact, according to the leaked document,

as many as 357 sensors in the Chicago metropolitan area

were broken, unreliable, or out of service at the time.

That's 9% of the total sensors in the city.

2,680 of SoundThinking sensors nationwide, one in 10,

were categorized as either broken, unreliable,

or out of service at the time the file was created,

allegedly late last year.

This past September, Mayor Brandon Johnson,

a vocal critic of ShotSpotter,

said the city will not renew

its contract with SoundThinking.

On a side note, one interesting use case for ShotSpotter

can be found way across the world in South Africa.

The Kruger National Park,

one of the most renowned wildlife conservation areas

in the world,

has sensors throughout its grounds,

presumably to curb the poaching of endangered species.

Wanna see if there's a ShotSpotter near you?

Click on the WIRED article in this video's description

and search the data yourself.

I'm Dhruv Mehrotra.

Thanks for watching On the Grid.

[mellow music]