Police Departments Reported ShotSpotter as Unreliable and Wasteful

ShotSpotter algorithmic systems locating gunshots were reported by police departments for containing high false positive rates and wasting police resources, prompting discontinuation.

Risk classification

• Primary risk domain: 7 AI system safety, failures, & limitations
• Primary risk subdomain: 7.3 Lack of capability or robustness

The system fails to perform reliably under varying environmental conditions, frequently misclassifying everyday noises like fireworks and car backfires as gunshots.

Additional risk subdomains

• 1.1 Unfair discrimination and misrepresentation: The system's deployment is concentrated in minority neighborhoods, resulting in disproportionate overpolicing and hostile law enforcement encounters for these groups.
• 5.1 Overreliance and unsafe use: Police departments rely heavily on the system's alerts, deploying officers on high alert expecting armed suspects based on unverified data.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The risks and harms arose from the AI system's failure to accurately classify sounds post-deployment, leading to unintended false alerts.

EU AI Act risk tier

• Risk tier: 2 High Risk

Risk Level 2: High Risk. The system is used in law enforcement applications, specifically for gunshot detection and dispatching police officers, which has significant implications for public safety and fundamental rights.

AI system and alleged parties

• AI purpose: Gunshot Detection; Smart Surveillance
• Behaviour type: Assistant
• Alleged developer: ShotSpotter
• Alleged deployer: Troy Police Department, Syracuse Police Department, San Francisco Police Department, San Antonio Police Department, New York City Police Department, Fall River Police Department, Chicago Police Department
• Alleged harmed parties: Troy residents, Troy Police Department, Syracuse residents, Syracuse Police Department, San Francisco residents, San Francisco Police Department, San Antonio residents, San Antonio Police Department, New York City residents, New York City Police Department, Fall River residents, Fall River Police Department, Chicago Residents, Chicago Police Department

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Negligible. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Substantial
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Substantial
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Substantial
• Civil rights: direct Negligible, indirect Minor
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Minor
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Negligible, indirect Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Physical

Reported: The report explicitly mentions the death of an unarmed 13-year-old in Chicago who was shot by officers responding to a ShotSpotter alert in 2021.

Indirectly caused: An unarmed 13-year-old, Adam Toledo, was shot and killed by Chicago police officers responding to a ShotSpotter alert in 2021.

Inferred additional harm: It is likely that other high-intensity police deployments triggered by false alerts have resulted in physical altercations, injuries, or excessive force, though specific numbers are not detailed.

Financial

Reported: The reports describe cities spending millions on ShotSpotter contracts, such as Chicago's $33 million contract and NYC's $54.6 million contract, representing significant public funds spent on an ineffective system.

Indirectly caused: Municipalities incurred substantial financial losses from annual subscription fees (e.g., $400,000 annually in Syracuse, $120,000 in Fall River) and wasted police labor costs.

Inferred additional harm: Millions of dollars in wasted public funds and police overtime pay spent investigating tens of thousands of false alarms across over 130 cities.

Differential treatment

Reported: The reports explicitly describe how the system is deployed almost exclusively in Black and Latinx communities, subjecting them to disproportionate police scrutiny.

Indirectly caused: Predominantly minority neighborhoods are subjected to thousands of unfounded, high-intensity police deployments, reinforcing racialized patterns of policing.

Inferred additional harm: Systemic discrimination where residents of color are disproportionately stopped, frisked, or subjected to protective pat-downs based on inflated gunshot statistics in their neighborhoods.

Civil rights

Reported: The reports describe civil liberties groups' concerns regarding overpolicing, wrongful stops, and potential voice surveillance violating privacy rights.

Indirectly caused: The system's false alerts lead to police officers conducting investigatory stops and protective pat-downs without proper reasonable suspicion, violating civil liberties.

Inferred additional harm: Numerous unquantified violations of Fourth Amendment rights against unreasonable searches and seizures due to officers using ShotSpotter history to justify stops.

Privacy

Reported: The reports mention concerns that ShotSpotter's always-on microphones can record private human conversations up to 50 feet away.

Indirectly caused: In at least one case, a California court accepted a voice recording captured by ShotSpotter microphones as evidence in a murder trial, demonstrating the capture of private conversations.

Inferred additional harm: Unconsented recording of outdoor conversations of thousands of residents living near the sensors, compromising general expectations of privacy.

Psychological

Reported: The reports describe residents in monitored neighborhoods experiencing heightened fear and distress due to aggressive police responses.

Indirectly caused: Residents in predominantly Black and Latinx neighborhoods face the psychological burden and distress of constant surveillance and high-alert police deployments.

Inferred additional harm: Widespread anxiety and trauma among community members, particularly youth, due to frequent high-intensity police presence and fear of wrongful confrontation.

Epistemic

Reported: The reports describe the company presenting misleading 97% accuracy claims that are not backed by independent scientific studies.

Indirectly caused: The propagation of unverified accuracy metrics by SoundThinking misleads city officials and the public about the system's effectiveness.

Inferred additional harm: Erosion of public trust in municipal decision-making and police data transparency due to the use of 'techwashed' statistics to justify policing resources.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 1000000

Potential causes

Management

• Sales-Driven Performance Claims: The 80% accuracy guarantee was created by marketing, not engineering.
• Prioritizing Growth Over Safety: Companies pushed system expansions while ignoring known accuracy issues.
• Poor Fiscal Stewardship: Cities spend millions on contracts despite low arrest and conviction rates.

Technology

• Algorithmic Inaccuracy: Software struggles to differentiate gunshots from fireworks or car backfires.
• Lack of Sensor Recalibration: Sensors in SF were not retested for a decade despite topography changes.
• Environmental Interference: Topography, building density, and wind degrade acoustic sensor accuracy.

Data Inputs

• Acoustic Noise Pollution: City noises like jackhammers and fireworks pollute the audio data streams.
• Missing Sensor Locations: Exact sensor locations are withheld from police, limiting data validation.
• Indoor Gunfire Exclusion: System is designed only for outdoor gunfire, failing to capture indoor shots.

Human Factors

• Subjective Human Review: Reviewers have broad discretion to overrule algorithmic classifications.
• Reviewer Bias and Manipulation: Reviewers modified gunshot locations and reports to suit police narratives.
• Police Officer Over-reliance: Officers respond on high alert, expecting armed suspects during false alarms.

Process and Methods

• Flawed Performance Metrics: Contracts measure performance using formulas that ignore false positives.
• Inadequate Data Collection: Police departments fail to track officer response times and false alarm costs.
• Lack of Scientific Testing: No independent, peer-reviewed studies exist to verify the 97% accuracy claim.

Regulatory Environment

• Lack of Oversight and Auditing: No federal or independent regulatory standards govern gunshot detection tools.
• Flawed Evidence Standards: Courts accept unverified acoustic reports as forensic evidence in trials.
• Privacy Regulation Gaps: Microphones can capture private conversations up to 50 feet away.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide extensive, detailed evidence from multiple independent audits (NYC Comptroller, MacArthur Justice Center), court testimonies, and municipal records across several cities. There is high consistency regarding the system's technical limitations, false positive rates, and social impacts, despite the manufacturer's disputed claims.
• Ambiguities identified: The exact proprietary algorithm used by ShotSpotter is not publicly disclosed, and the precise number of historical false positives that led to physical harm or wrongful arrests is not fully quantified.
• Alternative interpretations: The manufacturer argues that the lack of physical evidence at alert scenes does not prove a false positive, suggesting instead that casings were not found or guns did not eject them.

The deployment of the ShotSpotter AI gunshot detection system has led to significant local community impacts, including overpolicing in minority neighborhoods, civil rights concerns, and wasted municipal funds due to high false-alarm rates. From a national security perspective, the incident represents a minor, localized threat to societal stability and human rights, manageable within standard domestic legal and municipal frameworks.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. Represents an ongoing domestic policy and technology integration issue rather than an immediate national security crisis.
• Autonomy: Human-supervised. The AI system generates alerts that are reviewed by human operators before police dispatch, though high reliance is placed on the system.
• Novelty: Established threat. Acoustic sensor inaccuracies and algorithmic bias in policing are well-documented issues with existing precedents.

Impact by dimension

• Physical security: Negligible. Local law enforcement dispatch issues led to tragic localized incidents, but pose no threat to critical national infrastructure or national physical security.
• Information security: Negligible. No compromise of intelligence capabilities, classified models, or foreign information warfare operations indicated.
• Sovereignty: Negligible. Involves local municipal policing contracts and operations, with no threat to state authority, electoral systems, or federal government functions.
• Economic security: Negligible. Wasted municipal funds on ineffective technology do not threaten strategic industries, national economic stability, or technological competitive advantage.
• Societal stability: Minor. Systemic deployment in minority communities and high false-alarm rates contribute to overpolicing and privacy concerns, but the impact is localized and managed through municipal oversight.