This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Cruise autonomous vehicles experienced multiple incidents where they lost connectivity to company servers, causing the vehicles to become immobilized in active traffic lanes across San Francisco. These incidents resulted in significant traffic congestion and at least one documented instance where a vehicle blocked a fire truck responding to an emergency. Internal reports indicate that in some cases, the company lost the ability to track or remotely assist the vehicles, requiring manual intervention by staff to clear the roadways.
Cruise’s autonomous vehicles were shown on video stopping in the middle of the road and causing blockages in San Francisco, as they were disabled allegedly due to lost connection to their company’s server.
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.
Also classified as
The reports highlight regulatory challenges, including a state law loophole preventing moving violation citations for driverless cars and the CPUC approving commercial permits despite safety objections from local fire, police, and transit officials.
The Cruise vehicles failed to perform reliably under real-world conditions, repeatedly freezing and blocking traffic due to server connectivity losses and an inability to handle specific road scenarios.
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 incidents were caused by the Cruise autonomous vehicles experiencing unexpected technical failures and server connectivity losses after being deployed on public roads.
Multiple Cruise autonomous vehicles experienced server connectivity failures in San Francisco, leading to fleet immobilization that blocked active traffic lanes and obstructed an emergency vehicle. While highlighting safety vulnerabilities in autonomous transit systems, the incident has minor national security implications, limited to localized critical infrastructure disruption.
The Cruise vehicles operate independently as driverless robotaxis, navigating public streets and making real-time driving decisions based on environmental inputs without active human intervention, until they encounter technical failures.
High Risk: The AI system is deployed in critical road transport infrastructure. The reports describe vehicles blocking active traffic lanes and obstructing emergency services, which directly impacts public safety.
The primary goal of the Cruise vehicles is autonomous driving to transport passengers without a human driver, utilizing navigation assistance to route through San Francisco streets.
The analysis is supported by multiple detailed journalistic reports (Wired, SF Examiner, SFist) that cite internal Cruise communications, eyewitness accounts, and official letters from San Francisco city departments.
None, as the failure of the autonomous driving system and its real-world impact are clearly documented.
Technology
Loss of Server Connection
Vehicles lost contact with central server, triggering a safety freeze.
Inaccessible Backup System
Remote operators could not access fallback system to steer stopped cars.
Inability to Pull Over
Vehicles were not technologically capable of pulling over to the curb.
Data Inputs
Loss of Server Telemetry
Loss of connection cut off the essential data stream from central server.
Lack of Vehicle Location Data
Staff could not see where vehicles were located during the outage.
Lack of Passenger Communication
Staff could not communicate with passengers inside vehicles during outages.
Human Factors
Delayed Manual Recovery
Staff had to manually drive each vehicle to resolve physical blockades.
Whistleblower Safety Concerns
An employee raised concerns about regular contact loss and safety risks.
Emergency Responder Frustration
SFFD was highly concerned about response times blocked by stopped vehicles.
Process and Methods
Fail-Safe Defaults to Stopping
Vehicles are programmed to stop in place when encountering technical issues.
Ineffective Remote Operations
Fallback processes failed to let remote operators steer stopped vehicles.
Inadequate Incident Reporting
Cruise did not publicize the frequency of these fleet-wide outages.
Management
Prioritizing Commercialization
Rushing to launch paid services to compete with other autonomous vehicle firms.
Downplaying Systemic Failures
Management dismissed frequent fleet freezes as standard safety operations.
Inadequate Risk Assessment
Management failed to address known regular contact loss before scaling fleet.
Regulatory Environment
Premature Commercial Approval
CPUC approved driverless operations despite warnings from city officials.
Moving Violation Loophole
State law prevents police from citing driverless vehicles for moving violations.
Lack of Safety Enforcement
CPUC allowed operations to continue despite knowing about the blockages.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
