AI-enhanced surveillance and predictive policing tools are leading to wrongful arrests and traumatic police encounters because law enforcement often mistakes statistical probabilities for absolute, verified facts.
Key Points
- In October 2025, a Maryland teenager was detained at gunpoint after an AI camera misidentified a bag of chips as a weapon.
- Tennessee resident Angela Lipps spent five months in jail after facial recognition software incorrectly linked her to fraud crimes in another state.
- AI systems generate probabilistic risk scores rather than binary truths, yet these outputs are frequently treated as certainties by police departments.
- Confidence thresholds, which determine when an AI triggers an alert, are often set by vendors without public oversight or standardized legal requirements.
- Unlike judicial standards of proof, AI models lack the capacity to express uncertainty, often presenting incorrect predictions with high confidence.