This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
On February 14, 2016, a Google self-driving Lexus SUV operating in autonomous mode collided with a public bus in Mountain View, California. The vehicle attempted to merge into traffic to avoid sandbags but miscalculated the bus's intent to yield, resulting in a sideswipe collision. Google accepted partial responsibility for the incident, marking the first time one of its autonomous vehicles was found at fault in a crash.
On February 14, 2016, a Google autonomous test vehicle partially responsible for a low-speed collision with a bus on El Camino Real in Google’s hometown of Mountain View, CA.
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 human test driver failed to intervene despite seeing the bus, relying on the assumption that the AI system's prediction was correct.
The AI system failed to accurately predict the behavior of the public bus, demonstrating a lack of capability and robustness in handling complex real-world driving negotiations.
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 collision was caused by an incorrect prediction made by the AI system's software during autonomous operation on public roads post-deployment.
A Google self-driving Lexus SUV collided with a public bus at low speed in Mountain View, California, marking the first time an autonomous vehicle was found at fault. The incident resulted in minor property damage, no injuries, and carries negligible national security implications across all assessed categories.
The Google self-driving car was operating in autonomous mode, making real-time driving decisions such as lane positioning and predicting whether the bus would yield based on environmental sensor inputs.
High Risk: The AI system is used in critical infrastructure (traffic management and autonomous road transport), which has significant implications for public safety.
The AI system's primary function is to autonomously pilot the Lexus SUV on public roads, navigating traffic and obstacles without human intervention.
The reports provide consistent, detailed accounts of the collision on February 14, 2016, including official DMV accident reports, statements from Google, and details from the transit authority. There is high consensus on the facts of the primary incident.
None for the primary incident, as Google explicitly acknowledged partial responsibility for the bus collision.
Technology
Incorrect Yield Prediction
Software predicted the bus would yield because the car was ahead of it.
Flawed Lane Merging Algorithm
Algorithm failed to safely merge back into traffic after dodging sandbags.
Software Tweak Side Effects
Recent update to hug the right curb created complex lane-merging scenarios.
Data Inputs
Obstacle Detection (Sandbags)
Sandbags around a storm drain forced the vehicle to deviate from its lane.
Size-Based Yield Assumptions
System lacked deep understanding that large vehicles are less likely to yield.
Unexpected Road Obstacles
Sandbags around storm drain required sudden lane-merging maneuver.
Human Factors
Safety Driver Inaction
Test driver assumed the bus would slow down and did not take manual control.
Overreliance on Automation
Driver trusted the vehicle's decision-making instead of intervening early.
Bus Driver Expectation
Bus driver assumed the Lexus would stay stationary and did not yield.
Process and Methods
Inadequate Social Motoring Rules
System lacked logic for the informal give-and-take negotiation of driving.
Insufficient Edge Case Testing
Testing did not fully cover interactions with large municipal buses.
Public Road Software Testing
Testing unproven software tweaks directly on busy public roads.
Management
Aggressive Deployment Goals
Pushing for fully driverless vehicles before solving social driving norms.
Inadequate Risk Mitigation
Management failed to prevent unsafe assumptions in the lane-merging logic.
Regulatory Environment
Lack of Autonomous Standards
Absence of clear federal performance standards for self-driving decisions.
Permissive Testing Rules
California DMV allowed testing of autonomous modes with soft oversight.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
