This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A DUI arrest in Chicago was allegedly initiated based on a ShotSpotter gunshot detection alert. Public defenders challenged the reliability of the system and subpoenaed the company for data on its accuracy and analyst classification processes. ShotSpotter refused to comply with the discovery order, requesting a contempt-of-court order to facilitate an appeal, raising concerns about the transparency and overreliance on the technology by law enforcement.
A car stop resulting in a DUI arrest of its driver was allegedly based solely on a ShotSpotter alert, the reliability of which came into question by public defenders, who subpoenaed the company to assess its gunshot alert system.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Challenges in understanding or explaining the decision-making processes of AI systems, which can lead to mistrust, difficulty in enforcing compliance standards or holding relevant actors accountable for harms, and the inability to identify and correct errors.
Also classified as
The report highlights a civil lawsuit accusing the police department of overreliance on ShotSpotter alerts to justify stops and arrests.
The technology is criticized for being unreliable, frequently misidentifying fireworks or other ambient noises as gunshots.
The primary legal dispute centers on ShotSpotter's refusal to release documents explaining how its analysts and system assess and reclassify gunshot alerts, directly hindering judicial transparency.
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 stems from a ShotSpotter alert generated post-deployment, which unintentionally triggered police action and subsequent legal challenges regarding the system's reliability.
Added categories peak at 2: Minor (Psychological Harm, Indirect)
A domestic law enforcement incident in Chicago involving a DUI arrest based on a ShotSpotter gunshot detection alert. The company's refusal to comply with court subpoenas raises concerns about judicial transparency and civil rights, but the incident poses minor national security implications, as it is a localized legal and civil liberties issue handled through standard domestic judicial processes.
The report describes the AI system detecting acoustic events and flagging them for human analysts and police officers, who must then make the final decision to review, reclassify, or respond to the alert.
Risk Level 2: High Risk. The system is used in 'law enforcement applications' to detect gunshots and direct police response, which has significant implications for fundamental rights and public safety.
The system's primary function is to detect and localize gunshots using acoustic sensors, which also serves the broader goal of threat detection for law enforcement.
The report clearly outlines the legal dispute, the role of the ShotSpotter technology, and the specific harms (arrests, lack of transparency) associated with its deployment. The facts are consistent across the described cases.
The incident could be viewed strictly as a legal dispute over corporate proprietary data rather than an AI safety failure, though the underlying issue is the reliability of the AI tool.
Technology
Unreliable Acoustic Sensors
Sensors incorrectly identify fireworks and other noises as gunshots.
Retroactive Alert Modification
Analysts retroactively reclassify acoustic alerts days or weeks later.
Data Inputs
False Positive Audio Signals
Environmental noises like fireworks mimic gunshots, triggering false alerts.
Lack of Disclosed Error Data
Missing data on misidentified gunfire locations and missed gunshots.
Human Factors
Police Overreliance on Alerts
Police pulled over and arrested individuals based solely on automated alerts.
Subjective Analyst Decisions
Human analysts manually review and reclassify acoustic data in real-time.
Process and Methods
Secret Reclassification Methods
Proprietary methods used by analysts to reclassify alerts are kept secret.
Inadequate Alert Verification
Police stop citizens based on alerts without independent verification.
Management
Withholding Proprietary Data
Management chose contempt of court over releasing training and testing data.
Lack of Operational Transparency
Company refuses to disclose analyst qualifications and system error rates.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
