This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Tesla vehicles using vision-based Autopilot and FSD Beta systems have been subject to widespread reports of 'phantom braking,' where the vehicle suddenly and unexpectedly applies brakes without an actual hazard present. This issue, which surged following the removal of radar sensors in 2021, has led to over 750 complaints to the NHTSA, prompting a formal federal investigation into the safety of these driver-assistance systems.
In late 2021, Tesla owners’ complaints to the National Highway Traffic Safety Administration about sudden unexpected automatic braking rapidly increased, coinciding with when radar was no longer equipped in its Model 3 and Model Y vehicles.
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
Tesla's branding of 'Autopilot' and 'Full Self-Driving' can create a false sense of security, leading to overreliance on a system that still requires constant human vigilance.
The incident is a clear case of a lack of robustness and capability in the AI's perception system (Tesla Vision), which falsely detects non-existent hazards and triggers emergency braking.
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 phantom braking is caused by the deployed Tesla Autopilot/FSD AI system unexpectedly applying brakes due to false positives, which is an unintentional failure post-deployment.
Added categories peak at 2: Minor (Psychological Harm, Direct)
The incident involves a federal safety investigation into Tesla's vision-based Autopilot and FSD Beta systems following numerous reports of phantom braking. While presenting a physical safety risk on public roadways, the incident is a consumer product safety issue managed by standard regulatory processes (NHTSA) and has negligible implications for national security.
The reports describe Tesla's Autopilot and FSD systems operating independently to control the vehicle's speed and braking in real-time based on environmental inputs from cameras, without explicit human planning or intervention during the braking events.
High Risk: The system is used in critical transport infrastructure and safety-critical vehicle operations. The reports describe 'partially automated driving systems' that control braking and steering, which directly impacts physical safety on public roadways.
The primary function of Tesla's Autopilot and FSD Beta is autonomous driving and driver assistance, which includes threat detection to avoid collisions.
The reports provide consistent, detailed accounts of the phantom braking issue, backed by official NHTSA complaint statistics (758 complaints) and investigations. The technical transition to camera-only 'Tesla Vision' is well-documented as a primary context for the surge in issues.
Some might argue the issue is a standard mechanical/software defect rather than an AI safety issue, but the reliance on computer vision and deep learning for hazard detection makes it inherently an AI safety failure.
Technology
Removal of Radar Sensors
Transition to camera-only Tesla Vision removed critical sensor cross-checks.
Improper Decision Thresholds
System thresholds misclassify non-hazards like road trash as obstacles.
Camera Perception Vulnerabilities
Cameras struggle with environmental factors like shadows, glare, and rain.
Data Inputs
Single Sensor Data Stream
Relying only on cameras prevents cross-checking with radar or lidar data.
False Positive Triggers
Non-hazardous items like plastic bags or oncoming trucks trigger braking.
Environmental Noise
Glare, reflections, and bad weather degrade visual input data quality.
Human Factors
Bypassing Steering Pressure Sensors
Drivers easily bypass hands-on-wheel checks, leading to inattention.
False Sense of Security
Misleading names like Autopilot lead drivers to over-rely on the system.
Lack of Driver Intervention
Inattentive drivers fail to override sudden phantom braking events quickly.
Process and Methods
Beta Testing on Public Roads
Releasing unfinished FSD updates to customers to test in real-world driving.
Inadequate Edge Case Testing
Insufficient pre-release testing on metal bridges and S-shaped curves.
Flawed Software Rollouts
Updates like FSD 10.3 introduced severe phantom braking bugs.
Management
Prioritizing Rapid Deployment
Rushing software updates like FSD Beta to market ahead of robust validation.
Misleading System Marketing
Branding Level 2 systems as Autopilot and Full Self-Driving.
Disbanding PR Department
Eliminating public relations reduced transparency and communication on issues.
Regulatory Environment
Permissive Level 2 Regulations
Lack of strict rules on testing semi-autonomous systems on public roads.
Reactive Regulatory Oversight
NHTSA investigations began only after hundreds of owner complaints.
Lack of Pre-Market Certification
No mandatory safety certification required before deploying ADAS updates.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
