This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
ShotSpotter's acoustic gunshot detection system was used by Chicago police as critical evidence in a murder case, leading to the wrongful arrest and year-long incarceration of a 65-year-old man. Investigations revealed that the system frequently misclassifies sounds and that company employees manually modify alerts at the request of police, introducing human bias. The case highlights concerns regarding the lack of transparency, reliability, and potential for racial bias in algorithmic policing tools.
ShotSpotter audios were previously admitted to convict an innocent Black man in a murder case in Chicago, resulted in his nearly-one-year-long arrest before being dismissed by prosecutors as insufficient evidence.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
Also classified as
The system's sensors are deployed almost exclusively in Black and Latino neighborhoods, leading to systemic over-policing and discriminatory enforcement.
Police and prosecutors placed blind faith in the technology as direct evidence, ignoring disclaimers that it should only be used for initial investigative leads.
The core failure stems from the AI system's inability to reliably distinguish gunshots from other percussive sounds like fireworks or engine backfires, leading to highly inaccurate alerts.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The incident was primarily caused by the lack of capability and high false-positive rate of the deployed ShotSpotter AI system, which misclassified ambient noises as gunshots.
Added categories peak at 3: Substantial (Psychological Harm, Indirect)
The failure of ShotSpotter's acoustic gunshot detection system in Chicago led to a wrongful arrest and highlighted systemic issues with algorithmic policing tools. While causing localized civil liberties and human rights concerns, the incident has minor implications for national security.
The system acts as an assistant to law enforcement by detecting sounds, classifying them, and providing real-time alerts to dispatch officers, but it requires human analysts to verify the sounds and police officers to make final decisions.
High Risk: The system is used in law enforcement applications, specifically for crime detection and dispatching, which has significant implications for public safety and fundamental civil rights.
The system's primary designed function is to detect, classify, and locate the sounds of gunshots using acoustic sensors and AI algorithms.
The reports provide extensive, high-quality information including court documents, expert witness testimonies, nonpartisan watchdog investigations, and detailed case studies of affected individuals.
ShotSpotter maintains that its system is 97% accurate and that human analyst overrides are corrections rather than improper alterations.
The reports are highly sufficient for the classifications made.
Technology
Flawed Classification Algorithm
Algorithms misclassify fireworks or car backfires as gunshots.
Proprietary Black Box Software
No independent testing or peer review of the algorithm is allowed.
Data Inputs
Acoustic Noise Interference
Environmental sounds like fireworks and helicopters mimic gunshots.
Incorrect Initial Geolocation
System generated addresses up to a mile away from actual coordinates.
Human Factors
Analyst Manual Overrides
Analysts manually reclassified firecrackers to gunshots.
Police Tunnel Vision
Officers put blind faith in alerts and ignored other leads.
Lack of Formal Expert Training
No official training materials existed for forensic experts.
Process and Methods
Post-Hoc Report Alterations
Analysts modified alerts and shot counts at police request.
Misuse of Investigative Leads
Using alerts as primary trial evidence despite disclaimers.
Contaminated Feedback Loops
Training models with subjective or incorrect police feedback.
Management
Prioritizing Sales Over Accuracy
Accuracy guarantees were made by sales rather than engineers.
Secret Contract Extensions
City renewed contracts without public input or oversight.
Biased System Deployment
Sensors placed predominantly in minority neighborhoods.
Regulatory Environment
Lack of Independent Auditing
No regulatory requirement to audit forensic AI technology.
Absence of Standards for AI Evidence
Courts admit unvetted algorithmic evidence without validation.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
