← Back to incidents
Chinese AI Traffic Camera Fined Bus Advertisement as Jaywalker
MediumChinese AI traffic camera mistakenly identified face on bus advertisement as jaywalker, publicly shaming innocent person on violation display screen due to system's inability to distinguish between real faces and printed images.
Category
surveillance
Industry
Government
Status
Resolved
Date Occurred
Nov 1, 2018
Date Reported
Nov 22, 2018
Jurisdiction
China
AI Provider
Other/Unknown
Application Type
embedded
Harm Type
reputational
People Affected
1
Human Review in Place
No
Litigation Filed
No
facial_recognitiontraffic_enforcementchinasurveillancepublic_shamingliveness_detectiongovernment_ai
Full Description
In November 2018, an AI-powered traffic enforcement camera system in Ningbo, China, made a significant identification error that highlighted critical flaws in facial recognition technology deployment. The system was designed to automatically detect jaywalking violations and publicly display offenders' photos on large LED screens as part of China's social credit and public shaming enforcement mechanisms.
The incident occurred when a city bus drove past one of these smart traffic cameras. The bus featured a large advertisement on its side displaying a prominent face - reportedly that of Dong Mingzhu, chairwoman of major appliance manufacturer Gree Electric. The AI system's facial recognition algorithm detected this face from the advertisement and incorrectly classified it as a pedestrian committing a jaywalking violation.
Following its standard protocol, the system automatically processed what it believed to be a traffic violation. The individual's photo was extracted from the camera footage and immediately displayed on a nearby public LED screen designed to shame traffic violators. The display included the misidentified person's image along with violation details, effectively publicly accusing them of jaywalking when they were merely featured in a bus advertisement.
The error was discovered and reported by local media, drawing significant attention to the limitations of AI systems deployed in law enforcement contexts. The incident revealed that the facial recognition technology lacked basic liveness detection capabilities and could not distinguish between three-dimensional human faces and two-dimensional printed images. This fundamental flaw demonstrated insufficient testing and validation of the AI system before its deployment in a law enforcement capacity.
Local authorities acknowledged the error and removed the misidentified image from the public display. However, the incident had already gained widespread media coverage and social media attention, highlighting concerns about the reliability and appropriateness of automated law enforcement systems. The case became a notable example of AI system failures in China's expanding surveillance infrastructure and raised questions about the accuracy standards required for systems that can impact citizens' reputations and social credit scores.
Root Cause
The facial recognition system lacked ability to distinguish between real human faces and printed images on advertisements, treating 2D representations as live subjects for traffic violation detection.
Mitigation Analysis
Implementation of liveness detection algorithms could have prevented misidentification of static images. Human review protocols for flagged violations before public display would have caught the error. Additional training data including various 2D image formats and depth sensing technology could improve real-person detection accuracy.
Lessons Learned
This incident demonstrates the critical importance of implementing comprehensive testing protocols for AI systems deployed in law enforcement, particularly the need for liveness detection in facial recognition applications and human oversight mechanisms for automated public enforcement actions.
Sources
Chinese facial recognition system mistakes bus ad for famous CEO jaywalking
South China Morning Post · Nov 22, 2018 · news
China's facial recognition cameras catch bus ad jaywalking
BBC · Nov 22, 2018 · news