This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Apple's crash and fall detection software on iPhone 14 and Apple Watch models caused a significant surge in false 911 emergency calls, particularly in ski regions. The devices incorrectly identified high-intensity recreational activities as severe car crashes, overwhelming emergency dispatch centers with automated, false distress signals. This operational burden forced dispatchers to divert limited resources away from genuine emergencies, leading to frustration among emergency service personnel and prompting calls for software updates or user-side disabling of the feature.
Apple devices of skiers and snowboarders reportedly misclassified winter activities as accidents, which resulted in numerous false inadvertent distress calls to 911 dispatchers.
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.
The AI system failed to perform robustly under varying real-world conditions, specifically misinterpreting skiing and other high-intensity sports as vehicle crashes.
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 false emergency calls were caused by the automated decision-making of Apple's crash detection algorithm, which unintentionally misclassified recreational movements as car crashes after the devices were deployed.
Added categories peak at 2: Minor (Psychological Harm, Direct)
An unintentional AI software failure in Apple devices triggered widespread false 911 calls, temporarily straining local emergency services in North American ski regions. While causing operational friction for critical infrastructure, it posed negligible threats to national security.
The crash detection software operates independently in the background, continuously analyzing sensor data to make real-time decisions and trigger emergency calls without human intervention.
High Risk: The system is used as a safety component in consumer devices to dispatch emergency services, which has significant implications for public safety and critical infrastructure.
The primary goal of the system is to detect physical threats (car crashes and hard falls) using device sensors and automatically trigger emergency alerts.
The reports provide consistent, detailed accounts from multiple independent emergency dispatch centers and include statements from Apple, ensuring high reliability.
None
Technology
Overly Sensitive Algorithm
Software misinterprets high-intensity recreational movements as car crashes.
Inadequate Alert Notification
Users do not hear or feel alerts due to heavy winter clothing and helmets.
Automatic Satellite Calling
Devices automatically route false alarms through satellite links when offline.
Data Inputs
Misinterpreted Sensor Data
Bumping, bouncing, and rapid skiing deceleration mimic crash telemetry.
Lack of Environmental Context
Initial software lacked GPS context to filter out non-road locations.
Unresponsive User Input
Silence or lack of response is treated as confirmation of an emergency.
Human Factors
Inability to Detect Alerts
Skiers wearing helmets and layers fail to notice tactile and audio warnings.
Ignored Warning Vibrations
Some users feel vibrations but ignore them to avoid disrupting their activity.
Poor Device Placement
Placing devices in glove boxes hides warnings and increases jarring bumps.
Process and Methods
Inadequate Edge-Case Testing
Apple failed to test the software with extreme outdoor sports enthusiasts.
Flawed Default Settings
Fall detection is turned on automatically for users aged 55 and older.
Lack of Coordinated Dispatch
Dispatch centers lacked unified protocols to filter out automated ghost calls.
Management
Inadequate Risk Assessment
Management released the feature without assessing impact on emergency services.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
