Chicago police used ShotSpotter gunshot detection AI technology as key evidence to charge Michael Williams with murder, leading to his wrongful imprisonment for nearly a year before prosecutors dismissed the case due to insufficient evidence.
ShotSpotter is an AI-powered gunshot detection system that uses acoustic sensors and algorithms to identify and locate gunshots, deployed in over 100 cities including Chicago. The system initially detected a sound at 11:46 PM on May 31, 2020, classifying it as a firework with 98% confidence at a location about a mile from where prosecutors alleged a murder occurred. However, a ShotSpotter analyst manually reclassified the sound as a gunshot at the request of Chicago police, and the company later changed the reported location to match the prosecution's theory. This altered evidence became the centerpiece of the murder case against 65-year-old Michael Williams, who was charged with killing 25-year-old Safarian Herring. Williams spent 11 months in Cook County Jail, contracting COVID-19 twice and developing a hand tremor, before prosecutors dropped all charges in July 2021 citing insufficient evidence. Court documents reveal a pattern of ShotSpotter employees modifying alerts at police request, with the company's own expert witness testifying that such changes 'happen all the time.' Research shows that 89% of ShotSpotter alerts in Chicago lead to no evidence of gun crime, and the technology has been criticized for lacking independent scientific validation and transparency.
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.
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