Gunshot detection
Two different questions get called accuracy: whether the system hears gunfire, and whether police find anything when they arrive. The company answers the first and a city audit measured the second. Both answers are on the record, and they are not in conflict.
Updated
In one sentence
Networked microphones placed around a neighbourhood that listen for gunfire and report a location to police. The microphones are always on.
How the system works
Sensors go up on poles and rooftops across an area the agency has paid to cover. When several of them hear the same impulsive sound, the system compares arrival times and works out where it came from.
Software then decides whether the sound was a gunshot rather than a firework, a car door or a backfire, and in the leading product a human reviewer at the company checks the classification before an alert goes to dispatch. Officers are then sent to a point on a map, often with no caller, no complainant, and nobody at the scene expecting them.
What accuracy is measuring
This is where almost every public argument about the technology goes wrong, in both directions.
The company’s accuracy claim is about detection. Did the system hear the gunfire, classify it correctly, and put it in the right place? SoundThinking guarantees 90 percent in its contracts and publishes audited figures above 97 percent, with a false positive rate of half a percent.
The Chicago audit measured something else entirely. It asked what officers found when they arrived. Across seventeen months, 9.1 percent of alerts that generated a dispatch produced evidence of a gun-related offence.
Neither number refutes the other. A system can correctly detect a real gunshot and still send officers to a scene where there is nothing to find, because someone fired into the air, or the people involved left, or no crime is provable from an empty street. Anyone who quotes one figure to settle the argument is answering a question nobody asked.
What follows from that is the real question: what is a police dispatch worth when nine times in ten it produces nothing documented? That is a question about how a city spends its officers, and it is one a council can actually decide.
What Chicago decided
Chicago’s inspector general published its analysis in August 2021. Alongside the hit rate, it found that the presence of the system changed how officers behaved in the areas it covered, and raised concerns about alerts being used to justify stops.
The mayor announced in February 2024 that the city would not renew. Police stopped using it that September. The company said it had not been consulted on the decommissioning plan and had asked for more time.
The company also disputes the inspector general’s report directly, arguing that it measures investigative outcomes rather than detection performance. That objection is on the record and is a fair one to weigh, provided the reader understands that the outcome question is the one a city is paying for.
The microphones are always on
To detect an impulse, a sensor has to be listening all the time.
Speech captured by these sensors has been used as evidence in criminal cases. In an Oakland murder prosecution the system recorded a dying man naming his attacker. In a Massachusetts case from 2011 it recorded a street argument that preceded a fatal shooting. The company has said its microphones do not have voice recording capacity.
Both things are in the record: the company’s position, and two prosecutions in which recorded speech was used.
How to recognise one
The sensors are small weatherproof boxes mounted high on utility poles and building parapets, usually with a short antenna and no visible lens, and almost always unmarked. They are genuinely hard to identify from the ground, and mistaking a streetlight controller for one is easy.
The reliable route is the paperwork. These systems are bought by contract, cover an area measured in square miles, and appear in budgets and council agendas under the vendor’s name. We searched for a state statute specific to gunshot detection and could not find one. Where these systems have been stopped, it has generally been a city deciding not to renew.
Why agencies say they need it
-
Vendor claim
SoundThinking, which sells the ShotSpotter system, guarantees a 90 percent accuracy rate for detecting, locating and classifying gunfire in its customer agreements, and publishes an audit by an analytics firm reporting rates above 97 percent across its customers for 2019 to 2022 with a false positive rate of 0.5 percent. [3] [8]
-
Vendor claim
The company says an estimated 80 to 90 percent of gunfire is never reported to 911, so a detection system tells police about incidents that would otherwise produce no call at all. This is the strongest argument made for the technology and it is not addressed by the audits of what happened after alerts. [3]
What the concerns are
-
Most alerts did not lead to evidence of a gun crime
Chicago's Office of Inspector General examined every alert in the city between 1 January 2020 and 31 May 2021. Of the 41,830 alerts that led to a police dispatch, 4,556, or 9.1 percent, produced evidence of a gun-related criminal offence. SoundThinking disputes the report's framing, arguing that it measures investigative outcomes rather than whether the system correctly detected gunfire. [1] [2]
-
The audit also found the technology changed how police behaved
The inspector general's second finding was that the presence of the system altered police conduct in the areas it covered, and the office raised concerns about how alerts were being used to justify stops. This is a separate objection from the hit rate and does not depend on it. [1]
-
The microphones are always on, and recorded speech has reached a courtroom
The sensors listen continuously in order to detect an impulse. In an Oakland murder prosecution the system captured a dying man's words identifying his attacker, and the recording was used as evidence. In a 2011 Massachusetts case the sensors recorded a street argument that accompanied a fatal shooting, and that recording featured in the prosecution. The company has said its microphones do not have voice recording capacity. [6] [7]
-
A major customer ended its contract after reviewing the evidence
Chicago's mayor announced in February 2024 that the city would not renew, and the police department stopped using the system on 23 September 2024. The company said it had not been consulted on the decommissioning plan and had asked for a longer extension. [4] [5]
Where we have found it
37 counties documented so far. Each new county is added here as its research publishes.
- Alameda County, CA 30 community-reported devices
- Butte County, CA 1 also listed · 5 community-reported devices
- Contra Costa County, CA 1 system
- Active Gunshot detection [Antioch Police]
- Fresno County, CA 9 community-reported devices
- Humboldt County, CA 1 community-reported device
- Imperial County, CA 1 community-reported device
- Kern County, CA 74 community-reported devices
- Los Angeles County, CA 16 community-reported devices
- Mono County, CA 1 also listed
- Napa County, CA 1 community-reported device
- Nevada County, CA 1 community-reported device
- Orange County, CA 6 community-reported devices
- Placer County, CA 1 community-reported device
- Riverside County, CA 7 community-reported devices
- Sacramento County, CA 14 community-reported devices
- San Benito County, CA 1 system
- Active Gunshot detection and plate readers [Hollister Police]
- San Bernardino County, CA 13 community-reported devices
- San Diego County, CA 52 community-reported devices
- San Francisco County, CA 1 community-reported device
- San Mateo County, CA 2 community-reported devices
- Santa Clara County, CA 20 community-reported devices
- Siskiyou County, CA 1 community-reported device
- Solano County, CA 29 community-reported devices
- Sonoma County, CA 1 community-reported device
- Stanislaus County, CA 6 community-reported devices
- Tulare County, CA 2 community-reported devices
- Ventura County, CA 1 community-reported device
- Washoe County, NV 32 community-reported devices
- Multnomah County, OR 2 community-reported devices
- Yamhill County, OR 1 community-reported device
- Grays Harbor County, WA 1 community-reported device
- King County, WA 1 community-reported device
- Klickitat County, WA 1 community-reported device
- Lewis County, WA 3 community-reported devices
- Pierce County, WA 4 community-reported devices
- Spokane County, WA 1 community-reported device
- Yakima County, WA 4 community-reported devices
This list covers the counties we have researched, which is not the same as everywhere this technology operates. Counties we have not published say nothing either way.
Questions
- Does gunshot detection work?
- That depends which question you are asking, and the two are constantly merged. On detection, the company guarantees 90 percent accuracy contractually and publishes audited rates above 97 percent. On what happens next, Chicago's inspector general found that 9.1 percent of alerts that produced a police dispatch produced evidence of a gun-related offence. Both figures can be correct, because they measure different things. [1] [3] [8]
- Does ShotSpotter record conversations?
- The microphones listen continuously, and recorded speech has been used as evidence in at least two criminal cases, one in Oakland and one in Massachusetts. The company has said its microphones do not have voice recording capacity. [6] [7]
- Has any city stopped using it?
- Chicago did. The mayor announced in February 2024 that the contract would not be renewed, and the police department stopped using the system in September 2024. Other cities have debated or ended contracts, and each should be checked against that city's own records. [4] [5]
- Is there a law governing gunshot detection?
- We searched for a state statute specific to gunshot detection and could not find one. Where these systems have been restricted it has generally been by a city council decision about a contract rather than by a legislature. [4]
What you can do
Civic action
Purchases and policies for this technology are decided at public meetings any resident can attend and speak at. Start with your county's record.
Go to the Alameda County hub →Personal protection
Practical steps to shrink your personal data trail. Our guides publish at launch; until then, EFF's Surveillance Self-Defense is the best starting point.
Surveillance Self-Defense ↗Sources
- 1. Chicago Office of Inspector General, "The Chicago Police Department's Use of ShotSpotter Technology" · Primary document · · accessed · archived copy
- 2. SoundThinking, "Why the Chicago OIG Report is Misunderstood" · Primary document · accessed · archived copy
- 3. SoundThinking, "ShotSpotter Facts: separating fact from fiction in gunshot detection" · Primary document · accessed · archived copy
- 4.
- 5.
- 6.
- 7.
- 8. Edgeworth Economics, independent audit of ShotSpotter accuracy, 2019 to 2022 · Primary document · accessed · archived copy