Article 39 - FSD (Supervised) Dashboard

Overview
- A Gentle Introduction to FSD (Supervised)
- Definitions
Study 1: Mileage Distribution
Study 2: Surrogate Safety Metrics by Region
Study 3: Lane Change Safety Analysis
Study 4: Driver Distraction & Fatigue
- Part 1: Highway Fatigue Events (2025, North America)
Study 5: Fleet Collision Rates by Control Type and Road Class
Study 6: Engineering Fleet Testing
- Mileage Accumulated in European Countries
Study 7: Fixed Routes Testing
- Fixed Route Overview
- Fixed Route Test Results
Study 8: Max Speed Analysis
- Speed Comparison Analysis

Overview

This dashboard presents comprehensive safety evidence for FSD (Supervised) across multiple domains: collision rates comparing different control types in North America, safety performance indicators across regions including Europe, lane change safety analysis, quarterly collision trends, engineering fleet validation in European conditions, and structured fixed-route testing in 6 major European cities. The evidence demonstrates equivalent or superior safety performance across various metrics including collision rates, safety & environmental performance indicators, and behavioral competency through both statistical analysis and real-world validation, supporting Article 39, subparagraph 2, point (b) requirements for regulatory approval.

Methodological Note on Exposure Controls

The fleet-wide studies presented in this dashboard (Studies 1–6) stratify collision and surrogate safety rates by road class (highway, arterial, urban collector, local access), control type (FSD vs. manual with and without active safety), and time period (year and quarter). However, Tesla does not currently have the ability to directly stratify fleet-wide telemetry by all situational exposure variables that may influence collision risk. In particular, fleet-wide studies do not control for:

Traffic density and congestion levels at the time of driving
Weather and road surface conditions (e.g., rain, snow, wet roads)
Time of day and lighting conditions (e.g., daytime vs. nighttime driving)
Scenario complexity (e.g., frequency of unprotected turns, merges in heavy traffic, or construction zones)

These factors could influence where and when drivers choose to engage FSD (Supervised) versus drive manually, introducing potential selection effects into the observational comparison. Study 7 (Fixed Routes) partially addresses this limitation by evaluating FSD performance under controlled conditions across 6 European cities, with results stratified by weather conditions (clear, rain, etc.) and time of day (day, night, twilight, etc.) — but fixed routes represent structured validation rather than fleet-scale observational data.

Stratification by road class provides a meaningful proxy for driving environment complexity, as road classes correlate with typical traffic patterns, speed profiles, and intersection density. Nevertheless, residual confounding within road classes cannot be fully excluded. The statistical comparisons presented should be interpreted as observational associations under stratified exposure conditions, not as causal estimates from a randomized experiment.

Incompatibility with R171.01

The incompatibilities of FSD (Supervised) with R171.01 can be summarized as:

Neural Network Learning Approach: At the core of FSD (Supervised) is a neural network which learns patterns and behaviors that imitate high quality and safe driving from humans; natural emergent properties of this learning demonstrate good understanding and control for factors including but not limited to collision avoidance, anticipatory behavior, maintaining safe distances, traffic rules and signals, predictability, and smoothness. The driving trajectory is based on a continuous probabilistic optimization, making it incompatible with traditional rules-based, deterministic regulations.
Control Capability Restrictions: Many paragraphs of DCAS restrict the system in terms of control capabilities, and prevent FSD (Supervised) from performing intuitive and safe actions based on its learning from training data.
Boundary Condition Limitations: Many paragraphs in DCAS restrict the system in terms of boundary conditions, and limit usability and comfort in ways that prevent adoption of a system which provides a safety benefit for road safety.

Summary of Requested Exemptions

The requested exemptions from R171.01 for FSD (Supervised) can be summarized as:

Exemption Request 1 - FSD (Supervised) will perform system-initiated maneuvers and withhold HORs on highway roads, provided visual monitoring of the driver is possible.
Exemption Request 2 - FSD (Supervised) will perform system-initiated maneuvers and withhold HORs on non-highway roads, provided visual monitoring of the driver is possible.
Exemption Request 3 - System-initiated maneuvers from FSD (Supervised) will not be inhibited based on the detected driver state or presentation of driver disengagement warnings leading up to or at the time of initiation of the maneuver.
Exemption Request 4 - FSD (Supervised) will be allowed to induce lateral acceleration values beyond those specified in 5.3.7.1.2.
Exemption Request 5 - In certain cases, FSD (Supervised) will adjust the maximum speed to allow the system to drive the appropriate speed for the surrounding traffic conditions, even when this exceeds the detected speed limit.

Evidence Map: Studies Supporting Each Exemption

Studies 1, 2, 5, 6, and 7 establish the overall safety case for FSD (Supervised) and serve as foundational evidence across all exemption requests. The table below maps each exemption to the specific study parts that provide targeted evidence, with key findings summarized.

Overall System Safety Evidence

The following studies provide holistic safety evidence that underpins all exemption requests:

Study 1: Mileage Distribution — FSD (Supervised) accumulated nearly 2 billion miles in 2025 across all road classes, providing a statistically robust evidence base for all safety analyses.
Study 2: Surrogate Safety Metrics — Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) shows statistically significant improvements across every surrogate metric: 74-84% fewer AEB activations, 83-95% fewer harsh acceleration events, 55-75% fewer high lateral acceleration events, and higher blinker usage.
Study 5: Fleet Collision Rates — Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) demonstrates 40-69% lower major collision rates (IRR 0.31-0.60) and 48-59% lower minor collision rates (IRR 0.41-0.52) across all road classes. Compared to vehicles without active safety features, reductions are 71-90% (major IRR 0.10-0.29). Both with year-over-year improvement on non-highway roads.
Study 6: Engineering Fleet Testing — Over 793K FSD (Supervised) miles across 8 European countries with zero major or minor collisions observed due to FSD (Supervised) performance.
Study 7: Fixed Routes Testing — Over 230,000 scenario tests (each a unique interaction with a predefined road element such as a traffic light, roundabout, or crosswalk) across 6 European cities with zero safety-critical events and 99%+ overall pass rate.

Reading IRR values: IRR (Incidence Rate Ratio) compares the event rate of FSD (Supervised) to a baseline group. An IRR of 0.25 means FSD (Supervised)'s rate is 75% lower than the baseline (i.e., 1 − 0.25 = 0.75). IRR < 1 indicates FSD (Supervised) has a lower rate (safer); smaller values indicate larger safety improvements. See Definitions: Statistical Methods for full methodology.

Exemption	Study Evidence	Key Finding
Exemption 1 SIMs and HOR withholding on highways	Study 3, Part 1 Highway Lane Change Safety	Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) achieves a 99% reduction in minor collisions during highway lane changes (IRR 0.01).
	Study 3, Part 3 Lane Change Distraction States	Across 513M FSD (Supervised) lane changes in North America, only 2.5% had a DMS warning and 0.6% had an EOR request in the preceding 7 seconds, demonstrating high driver attentiveness during system-initiated maneuvers. This study was conducted using the North American DMS system.
	Study 4, Part 1 Highway Fatigue	Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) drivers experience 67% fewer fatigue events (IRR 0.333, p<0.001), demonstrating that driver monitoring via DMS remains effective when HORs are withheld.
	Study 2, Parts 1-2 AEB Rates	Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) shows 74-84% lower AEB activation rates across all road classes in NA (IRR 0.163-0.266) and 77-97% lower in Europe, indicating fewer imminent-collision scenarios across all driving contexts.
	Study 5, Parts 7-8 Regional Collision Comparison	Compared to European manually driven Tesla vehicles with active safety features, North American FSD (Supervised) outperforms across most road classes and collision types, supporting safety equivalence across diverse operating contexts.
Exemption 2 SIMs and HOR withholding on non-highway roads	Study 3, Part 2 Non-highway Lane Changes	Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) shows a 95% reduction in minor collisions (IRR 0.05), 89% reduction in FCW events (IRR 0.11), and 79% reduction in AEB events (IRR 0.21) across 214M non-highway lane changes.
	Study 3, Part 3 Lane Change Distraction States	Only 2.5% of FSD (Supervised) lane changes in North America had a DMS warning in the prior 7 seconds, showing drivers remain attentive during system-initiated maneuvers. This study was conducted using the North American DMS system.
	Study 5, Parts 1 & 4 Collision Rates by Road Class	Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) shows statistically significant lower major collision rates on all non-highway road classes: arterial 69% lower (IRR 0.31), urban collectors 50% lower (IRR 0.50), local access 40% lower (IRR 0.60). Compared to vehicles without active safety features, reductions are even larger: arterial 90% lower (IRR 0.10), urban collectors 74% lower (IRR 0.26), local access 71% lower (IRR 0.29).
	Study 1, Part 1 Mileage Distribution	Over 700M non-highway miles driven with FSD (Supervised) in 2025 in North America (400M+ urban collectors, 200M+ arterial, 80M+ local access), establishing sufficient exposure for non-highway safety validation.
	Study 7 Fixed Routes	99%+ pass rates across urban scenarios including roundabouts (98.9%), intersections (99.2%), crosswalks (99.7%), and cyclist yields (99.9%) in European cities.
Exemption 3 No SIM inhibition based on driver state	Study 3, Parts 1-2 Lane Change Safety	Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) achieves 82-99% improvement across all safety metrics (minor collisions, FCW, AEB) during lane changes on both highway and non-highway roads.
	Study 3, Part 3 Lane Change Distraction States	Across 513M FSD (Supervised) lane changes in North America, only 2.5% had a DMS warning and 0.6% had an EOR request in the preceding 7 seconds, demonstrating high driver attentiveness during system-initiated maneuvers. This study was conducted using the North American DMS system.
	Study 3, Part 4 Driver Situational Awareness	Across EU engineering vehicle lane changes, over 90% of drivers performed at least one relevant safety check (side mirror, side window, screen, or rearview mirror) before lane line encroachment, confirming that FSD (Supervised) does not induce complacency during system-initiated maneuvers.
	Study 4, Part 1 Highway Fatigue	Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) drivers experience one-third the fatigue events (IRR 0.333, p<0.001), indicating effective driver state monitoring.
	Study 7 Fixed Routes	Zero safety-critical events across over 230,000 scenario tests in 6 European cities, including complex system-initiated maneuvers.
Exemption 4 Lateral acceleration limits	Study 2, Part 5 Lateral Acceleration	Compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) produces 55-75% fewer lateral acceleration events ≥3 m/s² (IRR 0.252-0.447, p<0.001), meaning FSD (Supervised) already rarely exceeds the regulatory threshold and does so far less than human drivers.
Exemption 4 Lateral acceleration limits	Study 7 Fixed Routes (Roundabouts/Turns)	98.9% pass rate across 6,974 roundabout tests and zero safety-critical events through turns and complex maneuvers requiring lateral acceleration.
Exemption 5 Contextual speed determination	Study 8 Max Speed Analysis	Across 708 randomly selected European driving samples where the system exceeded the detected speed limit, FSD (Supervised) drove at or below the median surrounding traffic speed in 98% of high-speed cases and never exceeded the fastest surrounding vehicle in 99.9% of cases, demonstrating that FSD (Supervised) adjusts speed appropriately to match traffic flow rather than exceeding it.

A Gentle Introduction to FSD (Supervised)

Full Self-Driving (Supervised) as a Product

Since 2020, Tesla has developed a feature called Full Self-Driving (Supervised), an SAE Level 2 Advanced Driver Assistance System (ADAS) that supports drivers with the Dynamic Driving Task (DDT). FSD (Supervised) is able to maintain longitudinal and lateral control of the vehicle, under the supervision of an attentive driver, and perform functions on and off highway like: lane centering, making lane changes, taking highway exits, merging onto highways, handling traffic controls, navigating roundabouts, making turns at intersections, offsetting for objects, safely interact with vulnerable road users (e.g. pedestrians, bicyclists) and more. This feature is currently available in the United States of America, China, Canada, Mexico, Puerto Rico, Australia, New Zealand, and South Korea. Over 10.5 billion kilometers (6.5 billion miles) have been driven with the FSD (Supervised) feature active by Tesla owners on public roads across the globe. That is the equivalent of over 6,000 human lifetime's worth of driving or over 13,600 roundtrip visits to the Moon.

FSD (Supervised) uses eight high resolution cameras with a 360-degree field of view to observe the environment and a Tesla computer with custom silicon-chip to interpret the environment and navigate safely through it. It also uses additional sensors like IMUs, GPS, and in-cabin camera/torque sensors for additional functions like localization and driver monitoring. Based on deployment and usage data, FSD (Supervised) offers considerable safety enhancements to public roadways, both to drivers of Tesla vehicles and other users of the public roadway infrastructure (e.g. other vehicles, vulnerable road users). These safety enhancements span the spectrum of safety: fewer high-severity collisions like airbag deployments, fewer low-severity collisions like restraints control module (RCM) near-deploy events, fewer harsh braking events like AEBs, and even generally less unnecessary harsh accel/jerk control. As a result, Tesla is making concerted efforts to make this technology available to all Tesla owners globally, which requires exemption against a subset of regulations for an expansion into European markets.

For a deeper dive into FSD (Supervised) as a system, and to learn more about the end-to-end approach, please reference the presentation from Ashok Elluswamy, the Director of AI at Tesla, at the International Conference on Computer Vision.

The following video showcases a sample of training clips used to train FSD (Supervised). The system is trained on an extensive collection of driving scenarios, including edge cases and critical situations that most drivers may never encounter in their lifetime, ensuring robust performance across a wide range of real-world conditions.

Full Self-Driving (Supervised) as an SAE Level 2 System

The classification of FSD (Supervised) under the SAE J3016 Taxonomy of Driving Automation reflects the manufacturer's design intent for the division of roles and responsibilities between the driver and the system — not the system's technological capability. As codified in SAE J3016 Section 8.2, "the level assignment rather expresses the design intention for the feature." The level is fundamentally a delineation of responsibility.

FSD (Supervised) is designed and deployed as an SAE Level 2 system. The system provides sustained lateral and longitudinal vehicle motion control under the active, continuous supervision of the human driver. As specified in SAE J3016 Section 3.10, Level 2 driving automation "encompasses automation of... lateral and longitudinal vehicle motion control and limited OEDR associated with motion control." While the system perceives the driving environment to execute vehicle motion control, the driver retains full responsibility for supervising the driving environment, completing the Object and Event Detection and Response (OEDR) subtask, and intervening when necessary. This division of responsibility is the core of the system's safety philosophy and is reinforced through the vehicle's human-machine interface and driver education materials.

SAE J3016 Section 8.3 states that "the levels of driving automation do not specify or imply hierarchy in terms of relative merit, technology sophistication, or order of deployment." A Level 2 system may be technologically advanced and capable of navigating complex urban environments, but its classification remains Level 2 if the design intent requires the driver to supervise and assume responsibility for the OEDR subtask. FSD (Supervised) is precisely such a system — its advanced sensor suite, neural network processing, and ability to handle a wide range of driving scenarios do not alter its classification. The driver is always expected to monitor the environment, supervise the system, and be prepared to intervene when required.

This responsibility model is enforced through the Driver Monitoring System (DMS), a mandatory component of the Level 2 design. The DMS continuously monitors driver engagement through visual and torque-based inputs, issuing escalating warnings to ensure the driver remains attentive. In the event of prolonged inattention, the system initiates a "Take Over Immediately" sequence, followed by a controlled slowdown to a complete stop if the driver fails to respond. This is a safety protocol designed to enforce the driver's primary responsibility for the dynamic driving task — the defining characteristic of Level 2 automation.

The Safety Pillars of FSD (Supervised)

The safety of Tesla's Full Self-Driving (Supervised) system cannot be reduced to a singular evaluation of its performance in executing system-initiated maneuvers. While the system's ability to navigate complex environments is a critical component, it is only one pillar within a broader, multi-layered safety architecture designed to ensure the overall safety of the driver, passengers, and other road users. The system's safety is fundamentally underpinned by a holistic framework that includes robust driver monitoring, context-aware system limits, comprehensive driver education, and proactive user interface warnings. These elements work in concert to create a driving mode that is demonstrably safer than manual driving, even when individual performance metrics may not reflect perfection. The observed collision metrics, which show significantly lower rates of both major and minor incidents, are the result of this integrated safety approach. Factors such as the driver's ability to take over control, when necessary, the system's capacity to mitigate human error through consistent attention and predictable behavior, and the effectiveness of the Driver Monitoring System in maintaining driver engagement all contribute to the overall safety profile. Therefore, to evaluate the safety of FSD (Supervised) solely on its maneuver execution performance would be a fundamental mischaracterization of its safety design and would overlook the critical role played by the other pillars in ensuring a safe driving experience for all users.

This integrated safety framework is further substantiated by the detailed documentation provided in the supporting annexes, including the comprehensive Driver Monitoring System (DMS) report and the Driver Education materials. These documents outline the specific technical and procedural measures Tesla has implemented to ensure driver attentiveness and system controllability, which are essential for maintaining safety under SAE Level 2 operating conditions. The DMS, for instance, employs a multi-stage escalation protocol that adapts to driving context and driver behavior, ensuring that the driver is alerted appropriately when intervention is required. Similarly, the Driver Education program is designed to inform users of the system's capabilities and limitations, fostering a responsible and informed user base. By referencing these annexes, this submission demonstrates that the safety of FSD (Supervised) is not contingent on flawless performance but is instead the product of a carefully engineered ecosystem that leverages technology to augment, rather than replace, human oversight. This approach aligns with the core principle of SAE Level 2 automation, where the driver retains ultimate responsibility, and the system's design is intended to enhance, not diminish, the driver's ability to maintain safe control of the vehicle.

Video Clip Evidence of Behavioral and Safety Competency

The videos below demonstrates FSD (Supervised) operating across a diverse range of challenging driving scenarios, showcasing its behavioral and safety competency in real-world conditions. The compilation includes:

Last-second collision avoidance: Multiple instances where FSD makes critical decisions to prevent imminent collisions, demonstrating rapid perception and response capabilities
Adverse weather performance: Operation in rain, snow, and low-visibility conditions, showing robust decision-making across varied environmental conditions
Complex traffic interactions: Navigation through dense urban environments, highway merges, construction zones, and unpredictable behavior from other road users

These real-world examples complement the statistical evidence presented in the following studies, providing qualitative validation of FSD's safety performance and behavioral competency.

All videos have FSD (Supervised) engaged for the entirety of the video.

Note: Some videos may display a reconstruction of the in-car UI, some may not.

Definitions

Control Types

There are 3 primary control types used in this report, representing different levels of vehicle automation and safety feature availability.

Control Type	Description
Tesla Manual Driving (No Active Safety Features)	This represents Tesla vehicles that do not come equipped with active safety features, primarily early Model S/X vehicles that started selling in 2012. The average age of vehicles on European roads is 12.3 years, and the average age of vehicles on North American roads is 12.6 years (ACEA, 2025). This baseline provides the most representative comparison to the average vehicle population on public roads, as these older vehicles typically lack advanced driver assistance systems yet continue to accumulate significant mileage (120M miles tracked in Tesla's 2025 data). Note: Tesla does not log airbag near deployments or AEBs for vehicles without active safety features. This control type is removed for evidence presented around those cases.
Tesla Manual Driving (Active Safety Features)	Includes all Tesla active safety features such as Automatic Emergency Braking, Forward Collision Warning, Side Collision Warning, Obstacle Aware Acceleration, Blind Spot Monitoring, Lane Departure Avoidance, Pedal Misapplication Mitigation, and Emergency Lane Departure Avoidance. These active safety features demonstrate measurable safety improvements over the baseline vehicle population and are representative of the latest and most technologically advanced vehicles on the road.
FSD (Supervised, HW4)	FSD (Supervised) running on Hardware 4 vehicles, which represents the system configuration Tesla seeks to launch in the European market. This analysis controls specifically for HW4 to ensure the most representative performance data for regulatory evaluation.

Active Safety Features

Feature	Description
Automatic Emergency Braking (AEB)	Detects cars or obstacles that the vehicle may impact and applies the brakes
Forward Collision Warning	Warns of impending collisions with slower moving or stationary vehicles
Side Collision Warning	Warns of potential collisions with obstacles alongside the vehicle
Obstacle Aware Acceleration	Automatically reduces acceleration when an obstacle is detected in front of the vehicle while driving at low speeds
Blind Spot Monitoring	Warns when a vehicle or obstacle is detected when changing lanes
Lane Departure Avoidance	Applies corrective steering to keep the vehicle in its current lane
Emergency Lane Departure Avoidance	Steers the vehicle back into its driving lane when it detects that the vehicle is departing its lane and there could be a collision
Driver Drowsiness Detection/Warning	Uses the cabin camera to detect driver drowsiness and issues visual and audible alerts when fatigued driving patterns are identified (see Study 4: Driver Distraction & Fatigue)

Note: Active safety features are designed to assist drivers but cannot respond in every situation. It is the driver's responsibility to stay alert, drive safely and always be in control of the vehicle.

Road Classification

The analysis presented below consists of multiple studies that divide data into different road classes based on functional classification and traffic characteristics. When we refer to "non-highway" roads, we mean the overall collection of arterial, collector, and local access roads. There may be some studies that do not divide into the functional road classifications and instead, group together into 'non-highway'. These functional road classifications align with FHWA definitions.

Road Type	Definition	Flow of Traffic
Highway	Motorway/Freeways Limited-access, high speed roads, dedicated to long-distance mobility. Typically divided by medians, and roads where pedestrians or cyclists are not allowed.	Continuous flow High volume, limited access, uninterrupted mobility
Arterial	Roads of national importance Important roads for through traffic used to travel within the country and main roads running parallel to motorways.	Continuous flow (high mobility) High volume 'through-traffic' with low friction (limited driveways / stops). Flow mimics that of highway roads.
Urban Collector	Local roads of high importance Roads of national importance within a settlement. Roads making settlements accessible and roads used to travel within a part of a settlement.	Mixed flow between low and high friction Functionally slow traffic within a settlement (e.g. downtowns). Frequent interactions with buses, parking, VRUs (Vulnerable Road Users).
Local Access	Local roads of minor importance Destination only roads. Roads that are only used to reach a certain address or destination. Typically under 30 km/h.	Intermittent / Restricted Flow Very low volume/speed. Traffic is limited to immediate needs (e.g. residential / service roads). Traffic exists only to arrive or depart. No continuity.

SAE Level 2 Definition

The following table, derived from SAE J3016, defines the roles and responsibilities for SAE Level 2 driving automation systems. FSD (Supervised) operates as an SAE Level 2 system, where the driver retains full responsibility for the Dynamic Driving Task (DDT) and must remain engaged at all times.

Role of User	Role of Driving Automation System
Driver (at all times): • Performs the remainder of the DDT not performed by the driving automation system • Supervises the driving automation system and intervenes as necessary to maintain operation of the vehicle • Determines whether/when engagement and disengagement of the driving automation system is appropriate • Immediately performs the entire DDT whenever required or desired	Driving automation system (while engaged): • Performs part of the DDT by executing both the lateral and the longitudinal vehicle motion control subtasks • Disengages immediately upon driver request

Statistical Methods (IRR, Odds Ratios)

Statistical Methods: IRR (Incidence Rate Ratio) is a statistical measure comparing the rate of events between two groups. IRR < 1 indicates a lower rate in the comparison group (reduction), while IRR > 1 indicates a higher rate (increase). For collision and AEB rates, we use Poisson regression. For proportion comparisons (e.g., hands detection rates), we use either a two-proportion z-test (for large samples: all cell counts ≥5 and total sample ≥30) or Fisher's exact test (for small samples or sparse data: any cell count <5 or total sample <30). These tests produce Odds Ratios (OR) where OR > 1 indicates higher proportion in the comparison group. Fisher's exact test with Haldane-Anscombe correction provides more accurate confidence intervals when dealing with sparse data or small cell counts.

Statistical Significance Levels:

*** p<0.001: Highly significant
** p<0.01: Very significant
* p<0.05: Significant
† p<0.10: Approaching significance

Abbreviations

The following abbreviations are used throughout this dashboard, particularly in chart labels on smaller screens:

Abbreviation	Full Form
FSD(S, HW4)	FSD (Supervised, HW4)
Manual (AS)	Manual (With Active Safety Features)
Manual (No AS)	Manual (No Active Safety Features)
NA	North America
EU	Europe
Urb. Coll.	Urban Collectors
Local	Local Access
Non-hwy	Non-highway
AS	Active Safety (Features)
IRR	Incidence Rate Ratio

Study 1: Mileage Distribution

Understanding the distribution of fleet mileage across road classes is essential for demonstrating that FSD (Supervised) has accumulated sufficient real-world driving exposure to validate safety performance across all driving environments, from highways to complex urban and local access roads. This section establishes that the mileage base underlying the safety analyses in subsequent studies is broadly distributed, not concentrated on a single road type, and that FSD (Supervised) has been proven in use across the full range of road classes encountered in normal driving.

Data is presented for both North America (customer fleet) and Europe (engineering fleet for FSD, 2025 for other control types).

Exemption Relevance: This study contributes to the overall safety case underlying all exemption requests by establishing sufficient mileage exposure across all road classes. It is also directly relevant to Exemption Request 2 (SIMs and HOR withholding on non-highway roads), demonstrating over 700M non-highway FSD (Supervised) miles in 2025.

Study Takeaways

FSD (Supervised) has accumulated substantial mileage across all road classes in 2025, including hundreds of millions of non-highway miles in North America, providing a statistically robust basis for safety analysis across all driving environments.
Non-highway mileage (arterial, urban collector, and local access roads) constitutes a significant and growing share of FSD (Supervised) driving, demonstrating that the system is proven in use across the full spectrum of road types, not only highways.
FSD (Supervised) mileage grew approximately 4.5x from 2024 to 2025, with the largest proportional increases on non-highway road classes, further strengthening the evidence base for non-highway safety analyses.
The mileage distribution of manually driven Tesla vehicles remains stable across all road classes, providing a consistent and reliable baseline for rate-ratio comparisons.

Part 1: Mileage Distribution by Road Class (2025, North America)

The charts below outline the percentage of total miles driven by each control type on each road class in North America for 2025. This provides context on where and how FSD (Supervised) is being used relative to other control types in the Tesla fleet.

Part 1 Key Takeaways:

While FSD (Supervised) has proportionally more highway miles, there is substantial mileage across all non-highway road classes in 2025, including over 400 million miles on urban collectors, over 200 million on arterials, and over 80 million on local access roads in North America alone. This breadth of exposure ensures the safety analyses that follow are grounded in sufficient data across all road types.

View / Download Data

Mileage Breakdown Data

Road Class	Control Type	Total Miles	% of Control Type Miles
Highway	Manual (No Active Safety Features)	69,999,597	43.59%
Arterial	Manual (No Active Safety Features)	16,547,882	10.30%
Urban Collectors	Manual (No Active Safety Features)	54,623,024	34.01%
Local Access	Manual (No Active Safety Features)	19,432,632	12.10%
Highway	Manual (With Active Safety Features)	11,409,102,069	45.66%
Arterial	Manual (With Active Safety Features)	2,510,828,318	10.05%
Urban Collectors	Manual (With Active Safety Features)	8,258,191,430	33.05%
Local Access	Manual (With Active Safety Features)	2,811,665,301	11.25%
Highway	FSD (Supervised, HW4)	1,205,981,745	63.07%
Arterial	FSD (Supervised, HW4)	201,377,865	10.53%
Urban Collectors	FSD (Supervised, HW4)	422,593,313	22.10%
Local Access	FSD (Supervised, HW4)	82,110,744	4.29%

Part 2: Mileage Distribution - Year over Year (2024 vs 2025, North America)

Year-over-year comparison of mileage distribution. Each chart shows the percentage of each control type's total miles that occurred on each road class for 2024 and 2025.

Part 2 Key Takeaways:

FSD (Supervised) non-highway mileage share has increased significantly from 2024 to 2025, particularly on collector and local access roads, strengthening the evidence base for safety performance in complex driving environments. Tesla attributes this shift to the major software release of version 13 at the end of 2024.
All other control types show relatively stable mileage distributions year over year, with only minor fluctuations across road classes.

View / Download Data

Year	Control Type	Road Class	Total Miles	% of Control Type Miles
2024	FSD (Supervised, HW4)	Arterial	41,477,679	9.82%
2024	FSD (Supervised, HW4)	Highway	288,359,299	68.30%
2024	FSD (Supervised, HW4)	Local Access	13,474,332	3.19%
2024	FSD (Supervised, HW4)	Urban Collectors	78,889,226	18.69%
2024	Manual (No Active Safety Features)	Arterial	21,775,275	12.26%
2024	Manual (No Active Safety Features)	Highway	77,875,036	43.86%
2024	Manual (No Active Safety Features)	Local Access	21,020,717	11.84%
2024	Manual (No Active Safety Features)	Urban Collectors	56,890,206	32.04%
2024	Manual (With Active Safety Features)	Arterial	2,159,458,804	10.30%
2024	Manual (With Active Safety Features)	Highway	9,613,517,815	45.87%
2024	Manual (With Active Safety Features)	Local Access	2,297,932,524	10.96%
2024	Manual (With Active Safety Features)	Urban Collectors	6,889,532,316	32.87%
2025	FSD (Supervised, HW4)	Arterial	201,376,061	10.53%
2025	FSD (Supervised, HW4)	Highway	1,205,969,779	63.07%
2025	FSD (Supervised, HW4)	Local Access	82,109,989	4.29%
2025	FSD (Supervised, HW4)	Urban Collectors	422,589,564	22.10%
2025	Manual (No Active Safety Features)	Arterial	16,547,760	10.30%
2025	Manual (No Active Safety Features)	Highway	69,999,077	43.59%
2025	Manual (No Active Safety Features)	Local Access	19,432,437	12.10%
2025	Manual (No Active Safety Features)	Urban Collectors	54,622,552	34.01%
2025	Manual (With Active Safety Features)	Arterial	2,510,807,340	10.05%
2025	Manual (With Active Safety Features)	Highway	11,409,005,511	45.66%
2025	Manual (With Active Safety Features)	Local Access	2,811,642,252	11.25%
2025	Manual (With Active Safety Features)	Urban Collectors	8,258,122,908	33.05%

Part 3: Mileage Distribution - Quarterly Trends (Q1 2024 to Q4 2025, North America)

Quarterly trends in mileage distribution. Each chart shows the percentage of each control type's total miles that occurred on each road class over time (Q1 2024 to Q4 2025).

Part 3 Key Takeaways:

The quarterly trend confirms sustained growth in FSD (Supervised) non-highway mileage across all road classes. This increasing coverage in complex driving environments ensures the safety data continues to grow in both volume and representativeness, particularly following the version 13 software release at the end of 2024.

View / Download Data

Quarter	Control Type	Road Class	Total Miles	% of Control Type Miles
2024 Q1	FSD (Supervised, HW4)	Arterial	3,411,176	9.12%
2024 Q1	FSD (Supervised, HW4)	Highway	27,376,549	73.21%
2024 Q1	FSD (Supervised, HW4)	Local Access	873,494	2.34%
2024 Q1	FSD (Supervised, HW4)	Urban Collectors	5,731,544	15.33%
2024 Q1	Manual (No Active Safety Features)	Arterial	6,386,118	12.36%
2024 Q1	Manual (No Active Safety Features)	Highway	22,604,521	43.77%
2024 Q1	Manual (No Active Safety Features)	Local Access	6,122,849	11.85%
2024 Q1	Manual (No Active Safety Features)	Urban Collectors	16,535,055	32.01%
2024 Q1	Manual (With Active Safety Features)	Arterial	586,134,025	10.40%
2024 Q1	Manual (With Active Safety Features)	Highway	2,575,974,646	45.72%
2024 Q1	Manual (With Active Safety Features)	Local Access	617,732,231	10.96%
2024 Q1	Manual (With Active Safety Features)	Urban Collectors	1,854,356,636	32.91%
2024 Q2	FSD (Supervised, HW4)	Arterial	14,277,759	9.58%
2024 Q2	FSD (Supervised, HW4)	Highway	103,310,832	69.34%
2024 Q2	FSD (Supervised, HW4)	Local Access	4,495,521	3.02%
2024 Q2	FSD (Supervised, HW4)	Urban Collectors	26,904,588	18.06%
2024 Q2	Manual (No Active Safety Features)	Arterial	6,842,888	12.46%
2024 Q2	Manual (No Active Safety Features)	Highway	24,096,126	43.89%
2024 Q2	Manual (No Active Safety Features)	Local Access	6,491,149	11.82%
2024 Q2	Manual (No Active Safety Features)	Urban Collectors	17,472,607	31.82%
2024 Q2	Manual (With Active Safety Features)	Arterial	648,324,396	10.28%
2024 Q2	Manual (With Active Safety Features)	Highway	2,878,754,193	45.65%
2024 Q2	Manual (With Active Safety Features)	Local Access	697,531,663	11.06%
2024 Q2	Manual (With Active Safety Features)	Urban Collectors	2,081,089,348	33.00%
2024 Q3	FSD (Supervised, HW4)	Arterial	14,243,715	10.10%
2024 Q3	FSD (Supervised, HW4)	Highway	95,294,572	67.60%
2024 Q3	FSD (Supervised, HW4)	Local Access	4,648,292	3.30%
2024 Q3	FSD (Supervised, HW4)	Urban Collectors	26,778,600	19.00%
2024 Q3	Manual (No Active Safety Features)	Arterial	6,479,969	12.24%
2024 Q3	Manual (No Active Safety Features)	Highway	23,311,490	44.03%
2024 Q3	Manual (No Active Safety Features)	Local Access	6,223,617	11.76%
2024 Q3	Manual (No Active Safety Features)	Urban Collectors	16,924,941	31.97%
2024 Q3	Manual (With Active Safety Features)	Arterial	686,058,643	10.32%
2024 Q3	Manual (With Active Safety Features)	Highway	3,089,137,947	46.46%
2024 Q3	Manual (With Active Safety Features)	Local Access	713,506,020	10.73%
2024 Q3	Manual (With Active Safety Features)	Urban Collectors	2,160,574,546	32.49%
2024 Q4	FSD (Supervised, HW4)	Arterial	9,545,029	10.06%
2024 Q4	FSD (Supervised, HW4)	Highway	62,377,345	65.76%
2024 Q4	FSD (Supervised, HW4)	Local Access	3,457,025	3.64%
2024 Q4	FSD (Supervised, HW4)	Urban Collectors	19,474,494	20.53%
2024 Q4	Manual (No Active Safety Features)	Arterial	2,066,299	11.44%
2024 Q4	Manual (No Active Safety Features)	Highway	7,862,899	43.51%
2024 Q4	Manual (No Active Safety Features)	Local Access	2,183,102	12.08%
2024 Q4	Manual (No Active Safety Features)	Urban Collectors	5,957,603	32.97%
2024 Q4	Manual (With Active Safety Features)	Arterial	238,941,741	10.08%
2024 Q4	Manual (With Active Safety Features)	Highway	1,069,651,028	45.11%
2024 Q4	Manual (With Active Safety Features)	Local Access	269,162,609	11.35%
2024 Q4	Manual (With Active Safety Features)	Urban Collectors	793,511,786	33.46%
2025 Q1	FSD (Supervised, HW4)	Arterial	42,179,728	10.12%
2025 Q1	FSD (Supervised, HW4)	Highway	267,034,528	64.07%
2025 Q1	FSD (Supervised, HW4)	Local Access	16,880,617	4.05%
2025 Q1	FSD (Supervised, HW4)	Urban Collectors	90,668,273	21.76%
2025 Q1	Manual (No Active Safety Features)	Arterial	4,720,896	10.23%
2025 Q1	Manual (No Active Safety Features)	Highway	20,150,073	43.68%
2025 Q1	Manual (No Active Safety Features)	Local Access	5,576,228	12.09%
2025 Q1	Manual (No Active Safety Features)	Urban Collectors	15,685,748	34.00%
2025 Q1	Manual (With Active Safety Features)	Arterial	672,317,374	9.89%
2025 Q1	Manual (With Active Safety Features)	Highway	3,107,496,412	45.69%
2025 Q1	Manual (With Active Safety Features)	Local Access	763,515,310	11.23%
2025 Q1	Manual (With Active Safety Features)	Urban Collectors	2,257,817,954	33.20%
2025 Q2	FSD (Supervised, HW4)	Arterial	53,979,684	10.55%
2025 Q2	FSD (Supervised, HW4)	Highway	323,055,421	63.13%
2025 Q2	FSD (Supervised, HW4)	Local Access	21,981,171	4.30%
2025 Q2	FSD (Supervised, HW4)	Urban Collectors	112,706,136	22.02%
2025 Q2	Manual (No Active Safety Features)	Arterial	5,132,898	10.32%
2025 Q2	Manual (No Active Safety Features)	Highway	21,634,740	43.52%
2025 Q2	Manual (No Active Safety Features)	Local Access	6,035,960	12.14%
2025 Q2	Manual (No Active Safety Features)	Urban Collectors	16,913,961	34.02%
2025 Q2	Manual (With Active Safety Features)	Arterial	759,715,937	10.04%
2025 Q2	Manual (With Active Safety Features)	Highway	3,469,252,035	45.83%
2025 Q2	Manual (With Active Safety Features)	Local Access	845,614,939	11.17%
2025 Q2	Manual (With Active Safety Features)	Urban Collectors	2,494,722,258	32.96%
2025 Q3	FSD (Supervised, HW4)	Arterial	75,458,436	10.71%
2025 Q3	FSD (Supervised, HW4)	Highway	443,843,225	63.01%
2025 Q3	FSD (Supervised, HW4)	Local Access	30,267,254	4.30%
2025 Q3	FSD (Supervised, HW4)	Urban Collectors	154,840,393	21.98%
2025 Q3	Manual (No Active Safety Features)	Arterial	5,040,703	10.40%
2025 Q3	Manual (No Active Safety Features)	Highway	21,177,490	43.70%
2025 Q3	Manual (No Active Safety Features)	Local Access	5,824,321	12.02%
2025 Q3	Manual (No Active Safety Features)	Urban Collectors	16,415,285	33.88%
2025 Q3	Manual (With Active Safety Features)	Arterial	801,275,517	10.20%
2025 Q3	Manual (With Active Safety Features)	Highway	3,590,716,519	45.72%
2025 Q3	Manual (With Active Safety Features)	Local Access	880,925,849	11.22%
2025 Q3	Manual (With Active Safety Features)	Urban Collectors	2,581,186,425	32.86%
2025 Q4	FSD (Supervised, HW4)	Arterial	29,758,213	10.66%
2025 Q4	FSD (Supervised, HW4)	Highway	172,036,605	61.63%
2025 Q4	FSD (Supervised, HW4)	Local Access	12,980,947	4.65%
2025 Q4	FSD (Supervised, HW4)	Urban Collectors	64,374,763	23.06%
2025 Q4	Manual (No Active Safety Features)	Arterial	1,653,263	10.15%
2025 Q4	Manual (No Active Safety Features)	Highway	7,036,774	43.19%
2025 Q4	Manual (No Active Safety Features)	Local Access	1,995,929	12.25%
2025 Q4	Manual (No Active Safety Features)	Urban Collectors	5,607,559	34.42%
2025 Q4	Manual (With Active Safety Features)	Arterial	277,498,512	10.04%
2025 Q4	Manual (With Active Safety Features)	Highway	1,241,540,544	44.90%
2025 Q4	Manual (With Active Safety Features)	Local Access	321,586,154	11.63%
2025 Q4	Manual (With Active Safety Features)	Urban Collectors	924,396,271	33.43%

Part 4: Mileage Distribution by Road Class (2025, Europe)

The charts below outline the percentage of total miles driven by each control type on each road class in Europe for 2025. FSD (Supervised) data is from the engineering fleet, while other control types represent Tesla vehicles in the customer fleet.

Part 4 Key Takeaways:

The European engineering fleet demonstrates FSD (Supervised) coverage across all road classes in 2025, including arterial, urban collector, and local access roads. The distribution is broadly comparable to that of manually driven Tesla vehicles, confirming that the system has been exercised across the full range of European road types relevant to safety analysis.

View / Download Data

Mileage Breakdown Data

Road Class	Control Type	Total Miles	% of Control Type Miles
Highway	Manual (No Active Safety Features)	21,083,303	43.71%
Arterial	Manual (No Active Safety Features)	8,810,171	18.27%
Urban Collectors	Manual (No Active Safety Features)	13,337,683	27.65%
Local Access	Manual (No Active Safety Features)	5,003,909	10.37%
Highway	Manual (With Active Safety Features)	3,998,719,979	44.88%
Arterial	Manual (With Active Safety Features)	1,534,495,017	17.22%
Urban Collectors	Manual (With Active Safety Features)	2,445,492,725	27.45%
Local Access	Manual (With Active Safety Features)	931,675,279	10.46%
Highway	FSD (Supervised, HW4)	423,183	55.98%
Arterial	FSD (Supervised, HW4)	116,499	15.41%
Urban Collectors	FSD (Supervised, HW4)	174,624	23.10%
Local Access	FSD (Supervised, HW4)	41,616	5.51%

Part 5: Regional Comparison - Europe vs North America

Side-by-side comparison of mileage distribution between Europe and North America. For each control type, bars are grouped to show Europe (left) and North America (right), making regional differences immediately visible.

Part 5 Key Points:

European highway driving has a similar distribution of mileage across all control types compared to North America.
European non-highway driving has a different distribution of mileage across all control types (customers only), compared to North America.
- a ~70-80% increase on arterial road driving for manually driven Tesla vehicles. with less overall driving on collector and local access roads compared to North America.
- a ~20% decrease in collector/local access driving for manually driven Tesla vehicles.

View / Download Data

Region	Control Type	Road Class	Total Miles	% of Control Type Miles
North America	FSD (Supervised, HW4)	Arterial	201,377,865	10.53%
North America	FSD (Supervised, HW4)	Highway	1,205,981,745	63.07%
North America	FSD (Supervised, HW4)	Local Access	82,110,744	4.29%
North America	FSD (Supervised, HW4)	Urban Collectors	422,593,313	22.10%
North America	Manual (No Active Safety Features)	Arterial	16,547,882	10.30%
North America	Manual (No Active Safety Features)	Highway	69,999,597	43.59%
North America	Manual (No Active Safety Features)	Local Access	19,432,632	12.10%
North America	Manual (No Active Safety Features)	Urban Collectors	54,623,024	34.01%
North America	Manual (With Active Safety Features)	Arterial	2,510,828,318	10.05%
North America	Manual (With Active Safety Features)	Highway	11,409,102,069	45.66%
North America	Manual (With Active Safety Features)	Local Access	2,811,665,301	11.25%
North America	Manual (With Active Safety Features)	Urban Collectors	8,258,191,430	33.05%
Europe	FSD (Supervised, HW4)	Arterial	116,499	15.41%
Europe	FSD (Supervised, HW4)	Highway	423,183	55.98%
Europe	FSD (Supervised, HW4)	Local Access	41,616	5.51%
Europe	FSD (Supervised, HW4)	Urban Collectors	174,624	23.10%
Europe	Manual (No Active Safety Features)	Arterial	8,810,171	18.27%
Europe	Manual (No Active Safety Features)	Highway	21,083,303	43.71%
Europe	Manual (No Active Safety Features)	Local Access	5,003,909	10.37%
Europe	Manual (No Active Safety Features)	Urban Collectors	13,337,683	27.65%
Europe	Manual (With Active Safety Features)	Arterial	1,534,495,017	17.22%
Europe	Manual (With Active Safety Features)	Highway	3,998,719,979	44.88%
Europe	Manual (With Active Safety Features)	Local Access	931,675,279	10.46%
Europe	Manual (With Active Safety Features)	Urban Collectors	2,445,492,725	27.45%

Study 2: Surrogate Safety Metrics by Region

This study examines surrogate safety metrics across North America and Europe. We compare FSD (Supervised, HW4) surrogate safety to Manually driven Tesla vehicles (With Active Safety Features) as the established baseline. The surrogate safety measures are outlined below along with their reason for choice.

Data Period: Parts 1-5 (blinker, horn, and acceleration metrics) are from June 2 to September 9, 2025. Part 6 (AEB rates) represents data from January 1 to October 31, 2025.

Note: European FSD data is collected from engineering operators. Engineering operators are trained by Tesla to use the system with as little human input as possible. Due to this, surrogate safety metrics from the engineering fleet are appropriate to evaluate given the lack of driver input, making these metrics show the true performance of FSD (Supervised) on European roads.

Exemption Relevance: This study contributes to the overall safety case underlying all exemption requests. Part 5 (Lateral Acceleration) provides primary evidence for Exemption Request 4 (lateral acceleration limits), showing FSD (Supervised) produces 55-75% fewer lateral acceleration events ≥3 m/s² than manually driven Tesla vehicles with active safety features (IRR 0.252-0.447). Parts 1-2 (AEB Rates) support Exemption Requests 1 and 2, showing 74-97% lower AEB activation rates across all road classes.

Evidence-Based Rationale for Surrogate Safety Metrics

Surrogate Safety Metric	Evidence Based Rationale
AEB Activation Rate	Automatic Emergency Braking (AEB) is designed to intervene only in imminent collision scenarios. UNECE's AEBS regulation and NHTSA's new FMVSS 127 treat AEB as a last-resort collision-mitigation function activated when a crash is imminent. Numerous evaluations show that AEB substantially reduces rear-end crash frequency and severity when installed, which means each activation corresponds to a scenario that would otherwise have had a non-trivial crash risk. Therefore, a reduction in AEB rates can be linked to less potential near-miss scenarios. An IIHS study found that vehicles equipped with AEB had ~50% fewer police-reported rear-end crashes, and a similar reduction in rear-end crashes involving injuries. According to a 2025 analysis by Partnership for Analytics Research in Traffic Safety (PARTS), combining data across manufacturers and crash-reporting databases, AEB-equipped vehicles saw up to a 52% reduction in front-to-rear crashes. Because AEB intervenes only when a crash-likely trajectory is detected, each activation corresponds to a scenario with non-trivial crash risk. Therefore, a lower AEB activation rate, can reasonably be interpreted as reduced exposure to imminent-collision scenarios, which supports the use of lower AEB activation as a surrogate safety metric.
Longitudinal Acceleration and Deceleration Events (≥ 3 m/s²)	Harsh braking and rapid acceleration are long-established surrogate safety measures in naturalistic driving and telematics research. Hard braking thresholds around 3–3.5 m/s² are commonly used to identify elevated crash-risk situations, including late hazard detection, tailgating, and speed variability. For example, the Washington State Driving Behavior Analysis flags braking >3.2 m/s² as a high-risk maneuver. Surrogate-safety studies (e.g., Nikolaou et all. 2023) link high-magnitude longitudinal acceleration events to elevated frustration, instability in car-following behavior, and a higher probability of crash-contributing conflicts. Thus, reducing ≥3 m/s² events lowers exposure to situations associated with increased crash likelihood.
Lateral Acceleration Events (≥ 3 m/s²)	Naturalistic driving studies consistently show that hard cornering and high lateral-acceleration events are predictive of elevated crash and near-crash risk (Simons-Morton et al. 2012). In road-curve safety modeling, Ambros et al. (2019) calibrated lateral acceleration against real crash data and identified 0.3 g (≈3 m/s²) as a critical cutoff: curves with a higher proportion of observations above this threshold had significantly higher crash frequencies in the safety-performance models. Therefore, reducing the rate of ≥3 m/s² lateral-acceleration events corresponds to moving toward the lower-risk region of empirically validated crash-prediction models.
Horn Usage	Horn usage is widely used in driving-behavior research as a surrogate indicator of conflict, provocation, and aggressive responses. Behavioral scales such as the The Aggressive Driving Behavior Scale include honking frequency and intensity as validated aggression markers. Elevated honking rates in traffic correlate with acute stress, increased arousal, and frustration responses. Environmental-noise research (e.g., Munzel et al, 2017) shows that sudden high-intensity noise can trigger stress-physiological reactions, supporting the interpretation of excessive horn use as a marker of tense or unsafe traffic interactions.
Blinker Usage	Clear signaling reduces uncertainty and allows surrounding drivers and vulnerable road users to anticipate maneuvers. SAE field data estimate that turn-signal neglect contributes to millions of conflicts and crashes annually (Ponziani, 2012). Experimental studies show that proper indicator use improves other road users' ability to interpret intent, particularly for motorcyclists and pedestrians, reducing the likelihood of misjudgments that lead to conflicts (Micucci et al., 2019).

Study Takeaways

Analysis of large-scale operational data confirms that FSD (Supervised) exhibits a statistically and practically significant enhancement in key surrogate safety performance indicators across nearly every measured parameter when compared to manually driven Tesla vehicles.

A significantly lower AEB activation rate compared to manually driven Tesla vehicles. Since AEB intervenes only when a crash-likely trajectory is detected, each activation corresponds to a scenario with non-trivial crash risk. A lower rate indicates reduced exposure to imminent-collision scenarios.
Lower rates of both positive and negative high longitudinal-acceleration events (≥ 3 m/s²) per 1,000 miles, reflecting more stable acceleration and braking patterns. High-magnitude longitudinal events are associated with late hazard detection, tailgating, and unstable car-following behavior, all of which elevate crash likelihood.
Substantially fewer high lateral-acceleration events (≥ 3 m/s²) per 1,000 miles. Lateral acceleration above this threshold is a validated predictor of elevated crash and near-crash risk, making its reduction a meaningful indicator of smoother, safer driving.
Fewer horn events per 1,000 miles. While horn usage is commonly associated with aggressive driving, it is more broadly a marker of tense or unsafe traffic interactions. Elevated honking correlates with acute driver stress, frustration, and conflict with surrounding road users.
More blinker signals per 1,000 miles, indicating better signaling behavior and reduced turn-signal neglect. Proper indicator use reduces uncertainty for surrounding drivers and vulnerable road users, allowing them to better anticipate maneuvers and avoid conflicts.

For detailed research studies supporting the use of these surrogate safety metrics, please refer to the "Evidence-Based Rationale for Surrogate Safety Metrics" table above.

Part 1: AEB Rates by Control Type and Road Class (North America, 2025)

Automatic Emergency Braking (AEB) is a leading safety indicator that activates when the vehicle detects an imminent collision and applies brakes automatically. Vehicles without active safety features do not include AEB capability, so "Manual (No Active Safety Features)" is excluded from this analysis.

IRR Comparison - AEB Rates by Control Type and Road Class (North America, 2025)

Part 1 Key Points:

FSD (Supervised) demonstrates a statistically significant reduction in AEB activation rates compared to manually driven Tesla vehicles with active safety features, indicating fewer near-miss scenarios and enhanced overall safety performance.

View / Download Data

Chart Data

Control Type	Road Class	Total Events	Total Miles (M)	Rate per Million Miles	95% CI	Mean	Variance	Var/Mean Ratio	Overdispersed?
Manual (With Active Safety Features)	Highway	1,353,483	11,409.10	118.632	[118.432, 118.832]	4.5k	1.6M	349.733	Yes
Manual (With Active Safety Features)	Arterial	441,939	2,510.83	176.013	[175.495, 176.533]	1.5k	105.3k	72.399	Yes
Manual (With Active Safety Features)	Urban Collectors	1,909,337	8,258.19	231.205	[230.877, 231.533]	6.3k	2.1M	335.311	Yes
Manual (With Active Safety Features)	Local Access	966,676	2,811.67	343.809	[343.124, 344.495]	3.2k	243.1k	76.454	Yes
FSD (Supervised, HW4)	Highway	37,287	1,205.98	30.918	[30.605, 31.234]	122.655	1.6k	13.057	Yes
FSD (Supervised, HW4)	Arterial	6,930	201.38	34.413	[33.607, 35.233]	22.796	71.523	3.138	Yes
FSD (Supervised, HW4)	Urban Collectors	18,627	422.59	44.078	[43.447, 44.715]	61.273	513.110	8.374	Yes
FSD (Supervised, HW4)	Local Access	4,590	82.11	55.900	[54.295, 57.541]	15.099	101.878	6.747	Yes

Regression Results

Road Class	Comparison	FSD Rate	Baseline Rate	IRR	95% CI	Effect Size	Significant	Model Used
Highway	vs Manual (With Active Safety Features)	30.918	118.632	0.266	[0.255, 0.278]	-73.4%	Yes	Neg. Binomial
Arterial	vs Manual (With Active Safety Features)	34.413	176.013	0.198	[0.191, 0.205]	-80.2%	Yes	Neg. Binomial
Urban Collectors	vs Manual (With Active Safety Features)	44.078	231.205	0.192	[0.186, 0.198]	-80.8%	Yes	Neg. Binomial
Local Access	vs Manual (With Active Safety Features)	55.900	343.809	0.163	[0.158, 0.168]	-83.7%	Yes	Neg. Binomial

Part 2: AEB Rates by Control Type and Road Class (Europe, Engineering Fleet)

This section presents AEB activation rates comparing FSD (Supervised) engineering fleet operations to customer-driven manually operated Tesla vehicles with active safety features in Europe. This provides a direct comparison of AEB activation rates in European driving conditions.

As noted in the overview, European FSD data is collected from engineering operators. Engineering operators are trained by Tesla to use the system with as little human input as possible. Due to this, surrogate safety metrics from the engineering fleet are appropriate to evaluate given the lack of driver input, making these metrics show the true performance of FSD (Supervised) on European roads.

IRR Comparison - AEB Rates by Control Type and Road Class (Europe, Engineering Fleet)

Part 2 Key Points:

FSD (Supervised) engineering fleet operations in Europe demonstrate a statistically significant reduction in AEB activation rates compared to customer-driven manually operated Tesla vehicles with active safety features on highway, arterial, and urban collector road classes, consistent with North American findings and indicating fewer near-miss scenarios across diverse European road conditions.
On local access roads, the FSD AEB rate (682.7 per million miles) is marginally above the manual baseline (653.6 per million miles). This is based on 37 AEB events over 54,000 miles of FSD driving, with wide confidence intervals (480.7 to 941.1) that overlap with the manual rate. The limited mileage, small event count, and the inclusion of multiple software versions in the data (including the transition from V13 to V14 around October 2025) contribute to the imprecision of this estimate. On all other road classes, FSD AEB rates are 77% to 97% lower than the manual baseline.

View / Download Data

Chart Data

Control Type	Road Class	Total Events	Total Miles (M)	Rate per Million Miles	95% CI	Mean	Variance	Var/Mean Ratio	Overdispersed?
Manual (With Active Safety Features)	Highway	418,997	3,722.37	112.562	[112.221, 112.903]	1.0k	463.3k	459.970	Yes
Manual (With Active Safety Features)	Arterial	430,385	1,416.64	303.807	[302.900, 304.716]	1.0k	433.0k	420.528	Yes
Manual (With Active Safety Features)	Urban Collectors	1,031,523	2,270.17	454.382	[453.506, 455.260]	2.5k	2.6M	1050.497	Yes
Manual (With Active Safety Features)	Local Access	513,845	786.19	653.588	[651.802, 655.378]	1.2k	699.2k	579.663	Yes
FSD (Supervised, HW4)	Highway	4	0.53	7.593	[2.069, 19.442]	0.010	0.010	0.993	No
FSD (Supervised, HW4)	Arterial	1	0.15	6.703	[0.170, 37.344]	0.002	0.002	1.000	No
FSD (Supervised, HW4)	Urban Collectors	27	0.23	119.278	[78.605, 173.543]	0.066	0.115	1.753	No
FSD (Supervised, HW4)	Local Access	37	0.05	682.737	[480.710, 941.063]	0.090	0.565	6.277	Yes

Regression Results

Road Class	Comparison	FSD Rate	Baseline Rate	IRR	95% CI	Effect Size	p-value	Significant	Model Used
Highway	vs Manual (With Active Safety Features)	7.593	112.562	0.077	[0.029, 0.205]	-92.3%	0.0000 ***	Yes	Neg. Binomial
Arterial	vs Manual (With Active Safety Features)	6.703	303.807	0.025	[0.003, 0.176]	-97.5%	0.0002 ***	Yes	Neg. Binomial
Urban Collectors	vs Manual (With Active Safety Features)	119.278	454.382	0.297	[0.202, 0.437]	-70.3%	0.0000 ***	Yes	Neg. Binomial
Local Access	vs Manual (With Active Safety Features)	682.737	653.588	1.045	[nan, nan]	+4.5%	nan	No	Neg. Binomial
UNKNOWN_INVALID_SNA	vs Manual (With Active Safety Features)	0.000	109.472	0.000	[nan, nan]	-100.0%	nan	No	Neg. Binomial

Part 3: Positive Longitudinal Acceleration Events (≥ 3 m/s²)

The following section outlines positive longitudinal acceleration events (≥ 3 m/s²) per 1,000 miles broken down by region, road class, and control type. Lower bars indicate fewer positive longitudinal acceleration events (≥ 3 m/s²) per 1,000 miles, which is associated with more stable acceleration patterns, more predictable driving behavior, and less risky driving profiles.

As noted in the overview, European FSD data is collected from engineering operators. Engineering operators are trained by Tesla to use the system with as little human input as possible. Due to this, surrogate safety metrics from the engineering fleet are appropriate to evaluate given the lack of driver input, making these metrics show the true performance of FSD (Supervised) on European roads.

IRR Comparison - Positive Longitudinal Acceleration Events (≥ 3 m/s²)

Part 4 Key Points:

FSD (Supervised) usage shows a statistically significant decrease in longitudinal acceleration events above 3 m/s² on both European and North American roads compared to manually driven Tesla vehicles with active safety features, indicating more stable acceleration patterns and less risky driving profiles.

IRR values range from 0.051 to 0.172 with a p-value < 0.001 across all regions and road classes, indicating significantly lower longitudinal acceleration events above 3 m/s² for FSD (Supervised) compared to manually driven Tesla vehicles with active safety features.

View / Download Data

Rate Data

Region	Road Class	Control Type	Events	Total Miles	Rate per 1,000 miles	95% CI
North America	Highway	FSD (Supervised, HW4)	322,895	301,011,528	1.073	[1.069, 1.076]
North America	Highway	Manual (With Active Safety Features)	58,617,268	3,009,012,372	19.481	[19.476, 19.486]
North America	Non-highway	FSD (Supervised, HW4)	9,187,863	222,503,187	41.293	[41.266, 41.320]
North America	Non-highway	Manual (With Active Safety Features)	1,240,723,106	5,159,232,707	240.486	[240.473, 240.499]
Europe	Highway	Engineering - FSD (Supervised, HW4)	106	118,364	0.896	[0.733, 1.083]
Europe	Highway	Manual (With Active Safety Features)	31,794,845	1,810,939,301	17.557	[17.551, 17.563]
Europe	Non-highway	Engineering - FSD (Supervised, HW4)	1,426	130,511	10.926	[10.366, 11.508]
Europe	Non-highway	Manual (With Active Safety Features)	533,873,444	2,735,742,353	195.148	[195.131, 195.164]

Statistical Comparisons (FSD vs Manual)

Region	Road Class	FSD Rate	Manual Rate	IRR	95% CI	Effect Size	Significance
Europe	Highway	0.896	17.557	0.051	[0.042, 0.062]	-94.9%	significant
Europe	Non-highway	10.926	195.148	0.056	[0.053, 0.059]	-94.4%	significant
North America	Highway	1.073	19.481	0.055	[0.055, 0.055]	-94.5%	significant
North America	Non-highway	41.293	240.486	0.172	[0.172, 0.172]	-82.8%	significant

Part 4: Negative Longitudinal Acceleration Events (≥ 3 m/s²) - Braking

The following section outlines negative longitudinal acceleration events (≥ 3 m/s²) per 1,000 miles broken down by region, road class, and control type. Lower bars indicate fewer negative longitudinal acceleration events (≥ 3 m/s²) per 1,000 miles, which is associated with smoother and more predictable braking behavior, more comfortable rides, and fewer potential near-miss scenarios.

As noted in the overview, European FSD data is collected from engineering operators. Engineering operators are trained by Tesla to use the system with as little human input as possible. Due to this, surrogate safety metrics from the engineering fleet are appropriate to evaluate given the lack of driver input, making these metrics show the true performance of FSD (Supervised) on European roads.

IRR Comparison - Negative Longitudinal Acceleration Events (≥ 3 m/s²) - Braking

Part 5 Key Points: FSD (Supervised) usage shows a statistically significant decrease in braking events above 3 m/s² on both European and North American roads compared to manually driven Tesla vehicles with active safety features, indicating smoother and safer driving behavior with fewer abrupt braking events.

IRR values range from 0.141 to 0.417 with a p-value < 0.001 across all regions and road classes, indicating significantly lower braking events above 3 m/s² for FSD (Supervised) compared to manually driven Tesla vehicles with active safety features.

View / Download Data

Rate Data

Region	Road Class	Control Type	Events	Total Miles	Rate per 1,000 miles	95% CI
North America	Highway	FSD (Supervised, HW4)	231,683	301,011,528	0.770	[0.767, 0.773]
North America	Highway	Manual (With Active Safety Features)	5,550,561	3,009,012,372	1.845	[1.843, 1.846]
North America	Non-highway	FSD (Supervised, HW4)	679,452	222,503,187	3.054	[3.046, 3.061]
North America	Non-highway	Manual (With Active Safety Features)	84,141,099	5,159,232,707	16.309	[16.305, 16.312]
Europe	Highway	Engineering - FSD (Supervised, HW4)	24	118,364	0.203	[0.130, 0.302]
Europe	Highway	Manual (With Active Safety Features)	2,603,331	1,810,939,301	1.438	[1.436, 1.439]
Europe	Non-highway	Engineering - FSD (Supervised, HW4)	635	130,511	4.865	[4.494, 5.259]
Europe	Non-highway	Manual (With Active Safety Features)	42,635,800	2,735,742,353	15.585	[15.580, 15.589]

Statistical Comparisons (FSD vs Manual)

Region	Road Class	FSD Rate	Manual Rate	IRR	95% CI	Effect Size	Significance
Europe	Highway	0.203	1.438	0.141	[0.095, 0.210]	-85.9%	significant
Europe	Non-highway	4.865	15.585	0.312	[0.289, 0.337]	-68.8%	significant
North America	Highway	0.770	1.845	0.417	[0.416, 0.419]	-58.3%	significant
North America	Non-highway	3.054	16.309	0.187	[0.187, 0.188]	-81.3%	significant

Part 5: Lateral Acceleration Events (≥ 3 m/s²)

The following section outlines lateral acceleration events (≥ 3 m/s²) per 1,000 miles broken down by region, road class, and control type. Lower bars indicate fewer lateral acceleration events (≥ 3 m/s²) per 1,000 miles, which is associated with smoother driving and fewer potential near-miss scenarios.

As noted in the overview, European FSD data is collected from engineering operators. Engineering operators are trained by Tesla to use the system with as little human input as possible. Due to this, surrogate safety metrics from the engineering fleet are appropriate to evaluate given the lack of driver input, making these metrics show the true performance of FSD (Supervised) on European roads.

IRR Comparison - Lateral Acceleration Events (≥ 3 m/s²)

Part 3 Key Points:

FSD (Supervised) usage shows a statistically significant decrease in lateral acceleration events above 3 m/s² on both European and North American roads compared to manually driven Tesla vehicles with active safety features, indicating fewer potential near-miss scenarios and smoother driving behavior.

IRR values range from 0.252 to 0.447 with a p-value < 0.001 across all regions and road classes, indicating significantly lower lateral acceleration events above 3 m/s² for FSD (Supervised) compared to manually driven Tesla vehicles with active safety features.

View / Download Data

Rate Data

Region	Road Class	Control Type	Events	Total Miles	Rate per 1,000 miles	95% CI
North America	Highway	FSD (Supervised, HW4)	3,399,717	301,011,528	11.294	[11.282, 11.306]
North America	Highway	Manual (With Active Safety Features)	78,206,937	3,009,012,372	25.991	[25.985, 25.997]
North America	Non-highway	FSD (Supervised, HW4)	81,129,331	222,503,187	364.621	[364.542, 364.700]
North America	Non-highway	Manual (With Active Safety Features)	4,210,558,819	5,159,232,707	816.121	[816.096, 816.146]
Europe	Highway	Engineering - FSD (Supervised, HW4)	877	118,364	7.409	[6.927, 7.916]
Europe	Highway	Manual (With Active Safety Features)	53,397,347	1,810,939,301	29.486	[29.478, 29.494]
Europe	Non-highway	Engineering - FSD (Supervised, HW4)	39,693	130,511	304.135	[301.150, 307.142]
Europe	Non-highway	Manual (With Active Safety Features)	2,973,420,617	2,735,742,353	1086.879	[1086.840, 1086.918]

Statistical Comparisons (FSD vs Manual)

Region	Road Class	FSD Rate	Manual Rate	IRR	95% CI	Effect Size	Significance
Europe	Highway	7.409	29.486	0.251	[0.235, 0.268]	-74.9%	significant
Europe	Non-highway	304.135	1086.879	0.280	[0.277, 0.283]	-72.0%	significant
North America	Highway	11.294	25.991	0.435	[0.434, 0.435]	-56.5%	significant
North America	Non-highway	364.621	816.121	0.447	[0.447, 0.447]	-55.3%	significant

Part 6: Horn Usage

The following section outlines horn usage per 1,000 miles broken down by region, road class, and control type. Lower bars indicate fewer horn events per 1,000 miles, which is associated with less aggressive driving behavior and reduced noise pollution.

Note: Horn events are not automatically triggered by the vehicle and require driver input to activate. Nonetheless, horn usage is included as a surrogate safety metric to assess overall driving behavior and interaction with other road users.

As noted in the overview, European FSD data is collected from engineering operators. Engineering operators are trained by Tesla to use the system with as little human input as possible. Due to this, surrogate safety metrics from the engineering fleet are appropriate to evaluate given the lack of driver input, making these metrics show the true performance of FSD (Supervised) on European roads.

IRR Comparison - Horn Usage

Part 2 Key Points:

FSD (Supervised) usage shows a statistically significant decrease in horn usage on both European and North American roads compared to manually driven Tesla vehicles with active safety features, indicating potentially less aggressive driving behavior.

IRR values range from 0.131 to 0.322 with a p-value < 0.001 across all regions and road classes, indicating significantly lower horn usage for FSD (Supervised) compared to manually driven Tesla vehicles with active safety features.

View / Download Data

Rate Data

Region	Road Class	Control Type	Events	Total Miles	Rate per 1,000 miles	95% CI
North America	Highway	FSD (Supervised, HW4)	154,965	301,011,528	0.515	[0.512, 0.517]
North America	Highway	Manual (With Active Safety Features)	4,816,932	3,009,012,372	1.601	[1.599, 1.602]
North America	Non-highway	FSD (Supervised, HW4)	388,512	222,503,187	1.746	[1.741, 1.752]
North America	Non-highway	Manual (With Active Safety Features)	46,393,401	5,159,232,707	8.992	[8.990, 8.995]
Europe	Highway	Engineering - FSD (Supervised, HW4)	33	118,364	0.279	[0.192, 0.392]
Europe	Highway	Manual (With Active Safety Features)	2,587,926	1,810,939,301	1.429	[1.427, 1.431]
Europe	Non-highway	Engineering - FSD (Supervised, HW4)	154	130,511	1.180	[1.001, 1.382]
Europe	Non-highway	Manual (With Active Safety Features)	24,532,483	2,735,742,353	8.967	[8.964, 8.971]

Statistical Comparisons (FSD vs Manual)

Region	Road Class	FSD Rate	Manual Rate	IRR	95% CI	Effect Size	Significance
Europe	Highway	0.279	1.429	0.195	[0.139, 0.274]	-80.5%	significant
Europe	Non-highway	1.180	8.967	0.132	[0.112, 0.154]	-86.8%	significant
North America	Highway	0.515	1.601	0.322	[0.320, 0.323]	-67.8%	significant
North America	Non-highway	1.746	8.992	0.194	[0.194, 0.195]	-80.6%	significant

Part 7: Blinker Usage

The following section outlines blinker usage per 1,000 miles broken down by region, road class, and control type. Higher bars indicate more blinker events per 1,000 miles. For more comprehensive lane change study analysis, please reference Study 3, which goes over FCW/AEB/Minor collisions during a lane change, and driver distraction throughout lane change maneuvers.

As noted in the overview, European FSD data is collected from engineering operators. Engineering operators are trained by Tesla to use the system with as little human input as possible. Due to this, surrogate safety metrics from the engineering fleet are appropriate to evaluate given the lack of driver input, making these metrics show the true performance of FSD (Supervised) on European roads.

IRR Comparison - Blinker Usage

Part 1 Key Points:

FSD (Supervised) usage shows a statistically significant increase in blinker usage on European roads compared to manually driven Tesla vehicles with active safety features, indicating more consistent signaling behavior and lack of turn signal neglect.

Europe
- FSD (Supervised) engineering operations show that blinker usage is significantly higher than manually driven Tesla vehicles with active safety features across all road classes, indicating more consistent signaling behavior.
North America
- FSD (Supervised) blinker usage has a 1.071 IRR on highway, indicating slightly higher blinker events on highway, with 0.994 off-highway, indicating marginally less usage off-highway in North America. This could be due to the distribution of driving on and the differences to manually driven Tesla vehicles.

View / Download Data

Rate Data

Region	Road Class	Control Type	Events	Total Miles	Rate per 1,000 miles	95% CI
North America	Highway	FSD (Supervised, HW4)	233,413,850	301,011,528	775.432	[775.332, 775.531]
North America	Highway	Manual (With Active Safety Features)	2,178,390,116	3,009,012,372	723.955	[723.925, 723.986]
North America	Non-highway	FSD (Supervised, HW4)	369,086,851	222,503,187	1658.794	[1658.624, 1658.963]
North America	Non-highway	Manual (With Active Safety Features)	8,606,105,758	5,159,232,707	1668.098	[1668.063, 1668.133]
Europe	Highway	Engineering - FSD (Supervised, HW4)	122,174	118,364	1032.188	[1026.408, 1037.992]
Europe	Highway	Manual (With Active Safety Features)	1,509,724,961	1,810,939,301	833.670	[833.628, 833.712]
Europe	Non-highway	Engineering - FSD (Supervised, HW4)	303,776	130,511	2327.588	[2319.318, 2335.880]
Europe	Non-highway	Manual (With Active Safety Features)	4,287,459,503	2,735,742,353	1567.201	[1567.155, 1567.248]

Statistical Comparisons (FSD vs Manual)

Region	Road Class	FSD Rate	Manual Rate	IRR	95% CI	Effect Size	Significance
Europe	Highway	1032.188	833.670	1.238	[1.231, 1.245]	+23.8%	significant
Europe	Non-highway	2327.588	1567.201	1.485	[1.480, 1.490]	+48.5%	significant
North America	Highway	775.432	723.955	1.071	[1.071, 1.071]	+7.1%	significant
North America	Non-highway	1658.794	1668.098	0.994	[0.994, 0.995]	-0.6%	significant

Study 3: Lane Change Safety Analysis

This study compares safety-critical event rates during lane changes in North America, separated by highway and non-highway roads. The analysis examines three key metrics: Minor Collision events (minor collisions), Forward Collision Warning (FCW) events, and Automatic Emergency Braking (AEB) events per 10,000 lane changes. Vehicles without active safety features do not include AEB capability, so "Manual (No Active Safety Features)" is excluded from this analysis.

Statistical Model: Negative Binomial regression is used for all metrics in this study. Overdispersion testing (Pearson chi-squared dispersion ratio) on daily-aggregated lane change data showed dispersion ratios ranging from 3.0 to 156.0 across all metric and road type combinations, well above the threshold of 2.0, necessitating Negative Binomial over Poisson regression to account for the excess variance in event counts.

Lane Change Definition: A lane change is defined as a maneuver where the vehicle crosses a lane marking while the turn signal is active. All lane changes in this study occur where lane markings are present.

Trigger Condition: The trigger sources lane change scenarios based on the following conditions: (1) Vehicle eligibility: The trigger is deployed to all Hardware 4 vehicles with the FSD (Supervised) toggle enabled. This means that the "Manual (With Active Safety Features)" control group in this study consists exclusively of HW4 vehicles whose owners have opted in to FSD — the same vehicle population as FSD (Supervised). (2) Blinker activation: The vehicle's turn signal (blinker) is turned on, indicating an intent to change lanes. (3) Lane crossing: Ego is detected to have crossed the lane line.

Exemption Relevance: This study provides primary evidence for Exemption Requests 1, 2, and 3. Part 1 supports Exemption 1 (SIMs and HOR withholding on highways) with a 99% reduction in minor collisions during highway lane changes. Part 2 supports Exemption 2 (SIMs and HOR withholding on non-highway roads) with a 95% reduction in minor collisions across 214M non-highway lane changes. Part 3 supports Exemption 3 (no SIM inhibition based on driver state) by showing only 2.5% of lane changes had a DMS warning in the prior 7 seconds. Part 4 extends this with eye gaze analysis confirming that over 90% of drivers perform at least one relevant situational awareness check before encroachment.

Study Takeaways

Across over 377 million lane changes, FSD (Supervised) demonstrates statistically significant improvement across all surrogate safety metrics and all road classes presented during a lane change.

Additionally, driver distraction during system-initiated lane changes is exceptionally rare, with DMS warnings occurring in 2.49% of lane changes and EOR requests in 0.60% of lane changes (averaged across highway and non-highway roads). This indicates high driver engagement and attentiveness during FSD (Supervised) maneuvers.

Highway Lane Changes:

Minor Collision events: IRR 0.01 (95% CI: [0.005, 0.020]), p=<0.0001
FCW events: IRR 0.03 (95% CI: [0.025, 0.026]), p=<0.0001
AEB events: IRR 0.18 (95% CI: [0.169, 0.190]), p=<0.0001

Non-highway Lane Changes:

Minor Collision events: IRR 0.05 (95% CI: [0.038, 0.076]), p=<0.0001
FCW events: IRR 0.11 (95% CI: [0.112, 0.114]), p=<0.0001
AEB events: IRR 0.21 (95% CI: [0.198, 0.222]), p=<0.0001

Driver situational awareness analysis across 28,307 EU engineering vehicle lane changes (manual and FSD (Supervised) combined) confirms that FSD (Supervised) enhances rather than reduces situational awareness. FSD (Supervised) produces higher cumulative awareness than manual driving in both directions (left: 92.4% vs 90.7%; right: 91.8% vs 81.4%), providing strong evidence against overreliance or complacency.

Part 1: Highway Lane Change Safety Analysis

This analysis compares safety-critical event rates during lane changes on highway roads. Each row shows a specific metric (Minor Collision events, FCW, AEB) with the rate comparison chart on the left and the Incidence Rate Ratio (IRR) forest plot on the right.

Note: During the study period, there were zero major collisions during FSD (Supervised) lane changes on highway roads.

Data Period: October 6 to November 26, 2025.

Part 1 Key Takeaway

Across 163 million highway lane changes, FSD (Supervised) demonstrates statistically significant safety improvements across every surrogate safety metric presented:

Minor Collision events: IRR 0.01 (95% CI: [0.005, 0.020]), p=<0.0001
FCW events: IRR 0.03 (95% CI: [0.025, 0.026]), p=<0.0001
AEB events: IRR 0.18 (95% CI: [0.169, 0.190]), p=<0.0001

View / Download Data

Control Type	Lane Changes	Near Deployments	Near Deployments Rate (per 10k)	FCW Events	FCW Rate (per 10k)	AEB Events	AEB Rate (per 10k)
FSD (Supervised HW4)	163,035,595	8	0.0005	12567	0.77	1385	0.08
Manual (Active Safety)	105,933,655	528	0.05	323674	30.55	5024	0.47

Statistical Comparisons (FSD vs Manual)

Metric	IRR	95% CI	Significance
Near Deployments	0.010	[0.005, 0.020]	***
FCW Events	0.025	[0.025, 0.026]	***
AEB Events	0.179	[0.169, 0.190]	***

Part 2: Non-highway Lane Change Safety Analysis

This analysis compares safety-critical event rates during lane changes on non-highway roads. Each row shows a specific metric (Minor Collision events, FCW, AEB) with the rate comparison chart on the left and the Incidence Rate Ratio (IRR) forest plot on the right.

Note: During the study period, there were zero major collisions during FSD (Supervised) lane changes on non-highway roads.

Data Period: October 6 to November 26, 2025.

Part 2 Key Takeaway

Across 214 million non-highway lane changes, FSD (Supervised) demonstrates statistically significant safety improvements across every surrogate safety metric presented:

Minor Collision events: IRR 0.05 (95% CI: [0.038, 0.076]), p=<0.0001
FCW events: IRR 0.11 (95% CI: [0.112, 0.114]), p=<0.0001
AEB events: IRR 0.21 (95% CI: [0.198, 0.222]), p=<0.0001

View / Download Data

Control Type	Lane Changes	Near Deployments	Near Deployments Rate (per 10k)	FCW Events	FCW Rate (per 10k)	AEB Events	AEB Rate (per 10k)
FSD (Supervised HW4)	214,434,550	34	0.002	71239	3.32	1320	0.06
Manual (Active Safety)	393,721,865	1158	0.03	1154972	29.33	11561	0.29

Statistical Comparisons (FSD vs Manual)

Metric	IRR	95% CI	Significance
Near Deployments	0.054	[0.038, 0.076]	***
FCW Events	0.113	[0.112, 0.114]	***
AEB Events	0.210	[0.198, 0.222]	***

Part 3: Lane Change Distraction States

This section analyzes the total number of system initiated maneuvers (lane changes) completed in North America on FSD (Supervised, HW4), and what percentage of those have an Eyes On Road (EOR) request or DMS warning (any EOR or Hands On Request) in the 7 seconds preceding the lane change.

Data Period: October 6th to December 5th, 2025.

Part 3 Key Takeaway

Driver distraction during system-initiated lane changes is rare, indicating high driver engagement during FSD (Supervised) maneuvers:

Highway: - Only 2.16% of lane changes had a DMS warning in the preceding 7 seconds - Only 0.57% of lane changes had an EOR request in the preceding 7 seconds

Non-highway: - Only 2.81% of lane changes had a DMS warning in the preceding 7 seconds - Only 0.62% of lane changes had an EOR request in the preceding 7 seconds

These low percentages demonstrate that drivers remain highly attentive during system-initiated maneuvers across both highway and non-highway environments.

View / Download Data

Road Class	Total Lane Changes	With DMS Warning (7s before)	With EOR (7s before)	% with DMS	% with EOR
Non-highway	292,780,592.0	8,224,093.0	1,813,326.0	2.81%	0.62%
Highway	220,449,428.0	4,771,567.0	1,260,076.0	2.16%	0.57%

Study 4: Driver Distraction & Fatigue

FSD (Supervised) assists the driver with sustained lateral and longitudinal vehicle motion control, reducing the cognitive demands of driving. A less fatigued, more alert driver is better positioned to maintain situational awareness and intervene when necessary — a critical safety factor for any SAE Level 2 system. This study examines whether FSD (Supervised) usage is associated with lower driver fatigue levels compared to manual driving, using fatigue events detected by the Driver Monitoring System on highway roads.

Statistical Model: Poisson regression is used for the fatigue rate comparison, with the Incidence Rate Ratio (IRR) computed as the FSD rate divided by the manual rate and confidence intervals derived from the normal approximation for the log-IRR.

Exemption Relevance: This study provides primary evidence for Exemption Requests 1 and 3. Part 1 demonstrates that compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) drivers experience one-third the fatigue events (IRR 0.333, p<0.001), supporting Exemption 1 (SIMs and HOR withholding on highways) and Exemption 3 (no SIM inhibition based on driver state).

Study Takeaways

Highway fatigue analysis indicates that FSD (Supervised) usage is associated with significantly lower driver fatigue event rates compared to manually driven Tesla vehicles.

IRR of 0.333 (95% CI: [0.306, 0.362], p<0.001) demonstrates that drivers using FSD (Supervised) experience approximately one-third the fatigue events per 1,000 miles compared to manually driven Tesla vehicles with active safety features.

Part 1: Highway Fatigue Events (2025, North America)

View / Download Data

Driver Fatigue Data

Control Type	Miles	Fatigue Events	Events per 1,000 miles	95% CI Lower	95% CI Upper
Manual (With Active Safety Features)	2,053,356	2,744	1.34	1.29	1.39
FSD (Supervised, HW4)	1,575,803	701	0.44	0.41	0.48

Statistical Comparison (FSD vs Manual)

Road Type	Comparison	FSD Rate	Manual Rate	IRR	95% CI	p-value	Significant
Highway	FSD (Supervised, HW4) vs Manual (With Active Safety Features)	0.445	1.336	0.333	[0.306, 0.362]	0.0000 ***	Yes

Study 5: Fleet Collision Rates by Control Type and Road Class

This section presents collision rate comparisons between FSD (Supervised) and manually driven Tesla vehicles (both with and without active safety features) across all road classes. Regional comparisons to European driving conditions are also provided to contextualize performance metrics. Below are definitions and classifications used throughout this study.

Collision Definition

Tesla records collision events in accordance with 49 C.F.R. § 563.5, which defines a collision event as a crash or physical occurrence that causes any non-reversible deployable restraint to be deployed or that meets or exceeds the trigger threshold of a change in velocity (Delta-V) of 8 km/h occurring within a 150-millisecond period. Delta-V measures the change in a vehicle's velocity during a collision and is widely used in crash analysis to quantify impact severity. A higher Delta-V indicates greater forces acting on the vehicle and its occupants, making it a reliable indicator for discriminating between impact severities across a broad spectrum of collisions.

Consistent with 49 C.F.R. § 563.5, Tesla separates collisions into two categories:

Deployment collisions tend to have higher severity impacts where the vehicle's airbags or other non-reversible pyrotechnic restraints are deployed. Tesla refers to these collisions as "major collisions" in this report.

Non-deployment collisions tend to have lower severity impacts (Delta-V ≥ 8 km/h within 150 milliseconds) where neither airbags nor any other non-reversible pyrotechnic restraints are deployed. Tesla refers to these collisions as "minor collisions" in this report.

Collision Classification

If FSD (Supervised) was active at any point within five seconds leading up to a collision event, Tesla considers the collision to have occurred with FSD (Supervised) engaged for purposes of calculating collision rates. This approach accounts for the time required for drivers to recognize potential hazards and take manual control of the vehicle. This calculation ensures that our reported collision rates for FSD (Supervised) capture not only collisions that occur while the system is actively controlling the vehicle, but also scenarios where a driver may disengage the system or where the system aborts on its own shortly before impact.

Data Period

All collision and mileage data in this study covers January through October for both 2024 and 2025. The dataset was frozen as of October 2025 to undergo a comprehensive data validation and audit process conducted in collaboration with RDW, which spanned several months. This audit included verification of control type attribution, collision event deduplication, and confirmation against vehicle telemetry records. Using the same calendar months for both years ensures a consistent seasonal comparison window.

Exemption Relevance: This study contributes to the overall safety case underlying all exemption requests. It is directly relevant to Exemption Request 2 (SIMs and HOR withholding on non-highway roads): compared to manually driven Tesla vehicles with active safety features, FSD (Supervised) shows statistically significant lower collision rates on all non-highway road classes (arterial major IRR 0.31, urban collectors IRR 0.50, local access IRR 0.60). Compared to vehicles without active safety features, reductions are even larger (arterial IRR 0.10, urban collectors IRR 0.26, local access IRR 0.29). Parts 7-8 support Exemption Requests 1 and 2 by showing North American FSD (Supervised) outperforms European manual driving across most road classes.

Study Takeaways

FSD (Supervised) demonstrates statistically significantly lower major and minor collision rates compared to Manually driven Tesla vehicles (with and without active safety features) across all road classes in North America, indicating improved safety performance.
Temporal trends show statistically significant decreases in both major and minor collision rates for FSD (Supervised) over 2024-2025 across non-highway road classes,suggesting ongoing system improvements via software updates.
- While collision rates are trending downward for FSD (Supervised), Manually driven Tesla vehicles have not shown similar improvements.
- Collision rates are going down despite increased fleet mileage and decreased disengagement rates, indicating that FSD (Supervised) is becoming safer even as drivers rely on it more.
- Highway collision rates have slightly increased, however, exposure has increased by a large amount.
Regional comparisons show that European driving is statistically safer than North American driving for both major and minor collisions for all control types.
- Despite European roads being statistically safer to drive on, North American FSD (Supervised) still outperforms European Manual driving across all control types and road classes across all point estimates, with a majority showing statistical significance.
- For a direct comparison of NA to EU performance, please reference Study 2: Surrogate Safety Metrics.

Overdispersion Testing by Control Type and Road Class (North American Collisions)

Why Test for Overdispersion by Control Type and Road Class?

Following best practices in count data analysis (Cameron & Trivedi, 1990), we test each control type × road class combination separately for overdispersion. This approach recognizes that different driving conditions may have different variance characteristics.

Method: For each group, we calculate the variance-to-mean ratio from the daily collision counts. If ratio > 2, we use Negative Binomial regression for that comparison; otherwise, Poisson regression is appropriate.

Why this matters: Using the wrong model can lead to overly optimistic (too narrow) confidence intervals. By testing each group separately and using the appropriate model, we ensure our safety estimates are conservative and statistically valid.

Per-Group Overdispersion Test Results

Note: Mean and Variance are calculated from daily collision counts for each control type × road class combination over the study period.

Control Type	Road Class	N Days	Major Collisions				Minor Collisions
Control Type	Road Class	N Days	Mean	Variance	Var/Mean	Overdispersed?	Mean	Variance	Var/Mean	Overdispersed?
FSD (Supervised, HW4)	Arterial	304	0.095	0.100	1.05	No	0.385	0.422	1.10	No
FSD (Supervised, HW4)	Highway	304	0.467	0.487	1.04	No	1.586	1.854	1.17	No
FSD (Supervised, HW4)	Local Access	304	0.161	0.169	1.05	No	0.391	0.430	1.10	No
FSD (Supervised, HW4)	Urban Collectors	304	0.533	0.540	1.01	No	1.431	1.923	1.34	No
Manual (No Active Safety Features)	Arterial	304	0.076	0.077	1.01	No	0.000	0.000	0.00	No
Manual (No Active Safety Features)	Highway	304	0.148	0.140	0.94	No	0.000	0.000	0.00	No
Manual (No Active Safety Features)	Local Access	304	0.132	0.128	0.97	No	0.000	0.000	0.00	No
Manual (No Active Safety Features)	Urban Collectors	304	0.266	0.275	1.03	No	0.000	0.000	0.00	No
Manual (With Active Safety Features)	Arterial	304	3.793	3.960	1.04	No	11.740	14.912	1.27	No
Manual (With Active Safety Features)	Highway	304	7.993	29.267	3.66	Yes	31.411	92.626	2.95	Yes
Manual (With Active Safety Features)	Local Access	304	9.243	9.069	0.98	No	25.750	37.383	1.45	No
Manual (With Active Safety Features)	Urban Collectors	304	20.641	19.373	0.94	No	58.125	113.522	1.95	No

Model Selection Based on Overdispersion

For each pairwise comparison (e.g., FSD vs Manual on Highway), we check if either group shows overdispersion. If so, we use Negative Binomial regression for that comparison. Otherwise, we use Poisson regression.

This conservative approach ensures that statistical estimates account for any excess variability in the data, providing more reliable confidence intervals and p-values.

Part 1: Major Incidents per Million Miles (2025, North America)

IRR Comparison - Major Incidents per Million Miles (2025, North America)

Part 1 Key Takeaways

Significant decrease in major incidents per million miles when using FSD (Supervised) compared to all control types on all road classes.
- Comparisons to other control types show all 12 point estimates show IRRs below 1, with 10 of the 12 being statistically significant (p < 0.05).
Comparing to the average U.S. vehicle (vehicles without active safety features), IRR ranges from 0.104 to 0.29, all with p-values < 0.001.

View / Download Data

Chart Data

Control Type	Road Class	Total Events	Total Miles (M)	Rate per Million Miles	95% CI	Mean	Variance	Var/Mean Ratio	Overdispersed?
Manual (No Active Safety Features)	Highway	45	70.00	0.643	[0.469, 0.860]	0.148	0.140	0.944	No
Manual (No Active Safety Features)	Arterial	23	16.55	1.390	[0.881, 2.086]	0.076	0.077	1.015	No
Manual (No Active Safety Features)	Urban Collectors	81	54.62	1.483	[1.178, 1.843]	0.266	0.275	1.033	No
Manual (No Active Safety Features)	Local Access	40	19.43	2.058	[1.471, 2.803]	0.132	0.128	0.972	No
Manual (With Active Safety Features)	Highway	2,430	11,409.10	0.213	[0.205, 0.222]	7.993	29.267	3.661	Yes
Manual (With Active Safety Features)	Arterial	1,153	2,510.83	0.459	[0.433, 0.487]	3.793	3.960	1.044	No
Manual (With Active Safety Features)	Urban Collectors	6,275	8,258.19	0.760	[0.741, 0.779]	20.641	19.373	0.939	No
Manual (With Active Safety Features)	Local Access	2,810	2,811.67	0.999	[0.963, 1.037]	9.243	9.069	0.981	No
FSD (Supervised, HW4)	Highway	142	1,205.98	0.118	[0.099, 0.139]	0.467	0.487	1.043	No
FSD (Supervised, HW4)	Arterial	29	201.38	0.144	[0.096, 0.207]	0.095	0.100	1.046	No
FSD (Supervised, HW4)	Urban Collectors	162	422.59	0.383	[0.327, 0.447]	0.533	0.540	1.014	No
FSD (Supervised, HW4)	Local Access	49	82.11	0.597	[0.441, 0.789]	0.161	0.169	1.046	No

Regression Results

Road Class	Comparison	FSD Rate	Baseline Rate	IRR	95% CI	Effect Size	p-value	Significant	Model Used
Highway	vs Manual (No Active Safety Features)	0.118	0.643	0.183	[0.131, 0.256]	-81.7%	0.0000 ***	Yes	Poisson
Highway	vs Manual (With Active Safety Features)	0.118	0.213	0.550	[0.458, 0.661]	-45.0%	0.0000 ***	Yes	Neg. Binomial
Arterial	vs Manual (No Active Safety Features)	0.144	1.390	0.104	[0.060, 0.179]	-89.6%	0.0000 ***	Yes	Poisson
Arterial	vs Manual (With Active Safety Features)	0.144	0.459	0.314	[0.217, 0.453]	-68.6%	0.0000 ***	Yes	Poisson
Urban Collectors	vs Manual (No Active Safety Features)	0.383	1.483	0.259	[0.198, 0.338]	-74.1%	0.0000 ***	Yes	Poisson
Urban Collectors	vs Manual (With Active Safety Features)	0.383	0.760	0.505	[0.432, 0.590]	-49.5%	0.0000 ***	Yes	Poisson
Local Access	vs Manual (No Active Safety Features)	0.597	2.058	0.290	[0.191, 0.440]	-71.0%	0.0000 ***	Yes	Poisson
Local Access	vs Manual (With Active Safety Features)	0.597	0.999	0.597	[0.450, 0.792]	-40.3%	0.0003 ***	Yes	Poisson

Part 2: Major Incidents - Year over Year Comparison (2024 vs 2025, North America)

Year-over-Year Comparison: Each chart below represents one road class, showing all control types compared between 2024 and 2025. Both years are filtered to the same 10-month period for consistency (January to October).

Note on year-over-year variation: Small fluctuations in rate estimates between years are expected. In 2024, FSD accumulated significantly fewer miles, resulting in rate estimates with wider confidence intervals. Year-over-year changes where confidence intervals overlap or where the IRR is not statistically significant (p > 0.05) should not be interpreted as meaningful trends. The overall regression analysis in Part 1, based on the full 2025 dataset with substantially greater statistical power, provides the most reliable safety comparison.

IRR Comparison - Major Incidents - Year over Year Comparison (2024 vs 2025, North America)

Part 2 Key Takeaways

FSD (Supervised) has improved in terms of major collision rates year over year across every non-highway road class.
- This data, paired with disengagement rate improvements (see study 4) and total mileage increases (see study 1), indicate that FSD (Supervised) is becoming safer, more reliable, and drivers are becoming more comfortable using it.
Manually driven Tesla vehicles have not shown similar improvements.
On highways, the nominal 17% year-over-year increase in FSD major collision rates is not statistically significant (p=0.439, IRR 95% CI: 0.785 to 1.746, spanning both improvement and increase). The 2024 highway rate was based on 29 events over 288M miles, while the 2025 rate is based on 142 events over 1,206M miles, providing greater statistical power. Other control types also showed nominal increases on highways (Manual with Active Safety: +4.6%, Manual without Active Safety: +4.3%, both also not significant), suggesting this is not an FSD-specific trend, as all control types moved in the same direction on highways. FSD remains 45% safer than manually driven Tesla vehicles with active safety features on highways (IRR 0.55, p < 0.001) in the overall 2025 analysis.

View / Download Data

Year-over-Year Change Summary

Control Type	Road Class	2024 Events	2024 Miles (M)	2024 Rate	2024 95% CI	2025 Events	2025 Miles (M)	2025 Rate	2025 95% CI	Absolute Change	Percent Change	Mean	Variance	Var/Mean Ratio	Overdispersed?
Manual (No Active Safety Features)	Highway	48	77.88	0.616	[0.454, 0.817]	45	70.00	0.643	[0.469, 0.860]	+0.026	+4.3%	0.148	0.140	0.944	No
Manual (No Active Safety Features)	Arterial	31	21.78	1.424	[0.967, 2.021]	23	16.55	1.390	[0.881, 2.086]	-0.034	-2.4%	0.076	0.077	1.015	No
Manual (No Active Safety Features)	Urban Collectors	71	56.89	1.248	[0.975, 1.574]	81	54.62	1.483	[1.178, 1.843]	+0.235	+18.8%	0.266	0.275	1.033	No
Manual (No Active Safety Features)	Local Access	29	21.02	1.380	[0.924, 1.981]	40	19.43	2.058	[1.471, 2.803]	+0.679	+49.2%	0.132	0.128	0.972	No
Manual (With Active Safety Features)	Highway	2,065	9613.52	0.215	[0.206, 0.224]	2,430	11409.01	0.213	[0.205, 0.222]	-0.002	-0.8%	1.599	6.793	4.249	Yes
Manual (With Active Safety Features)	Arterial	1,101	2159.46	0.510	[0.480, 0.541]	1,153	2510.81	0.459	[0.433, 0.487]	-0.051	-9.9%	0.759	1.405	1.852	No
Manual (With Active Safety Features)	Urban Collectors	5,342	6889.53	0.775	[0.755, 0.796]	6,275	8258.12	0.760	[0.741, 0.779]	-0.016	-2.0%	4.128	23.825	5.771	Yes
Manual (With Active Safety Features)	Local Access	2,347	2297.93	1.021	[0.980, 1.064]	2,810	2811.64	0.999	[0.963, 1.037]	-0.022	-2.1%	1.849	5.238	2.834	Yes
FSD (Supervised, HW4)	Highway	29	288.36	0.101	[0.067, 0.144]	142	1205.97	0.118	[0.099, 0.139]	+0.017	+17.1%	0.467	0.487	1.043	No
FSD (Supervised, HW4)	Arterial	10	41.48	0.241	[0.116, 0.443]	29	201.38	0.144	[0.096, 0.207]	-0.097	-40.3%	0.095	0.100	1.046	No
FSD (Supervised, HW4)	Urban Collectors	53	78.89	0.672	[0.503, 0.879]	162	422.59	0.383	[0.327, 0.447]	-0.288	-42.9%	0.533	0.540	1.014	No
FSD (Supervised, HW4)	Local Access	14	13.47	1.039	[0.568, 1.743]	49	82.11	0.597	[0.441, 0.789]	-0.442	-42.6%	0.161	0.169	1.046	No

Regression Results

Manual (No Active Safety Features)

Road Class	Comparison	2024 Rate	2025 Rate	IRR	95% CI	Effect Size	p-value	Significant	Model Used
Highway	2025 vs 2024	0.616	0.643	1.043	[0.694, 1.566]	+4.3%	0.8393	No	Poisson
Arterial	2025 vs 2024	1.424	1.390	0.976	[0.569, 1.674]	-2.4%	0.9306	No	Poisson
Urban Collectors	2025 vs 2024	1.248	1.483	1.188	[0.864, 1.634]	+18.8%	0.2888	No	Poisson
Local Access	2025 vs 2024	1.380	2.058	1.492	[0.925, 2.406]	+49.2%	0.1009	No	Poisson

Manual (With Active Safety Features)

Road Class	Comparison	2024 Rate	2025 Rate	IRR	95% CI	Effect Size	p-value	Significant	Model Used
Highway	2025 vs 2024	0.215	0.213	1.046	[0.932, 1.174]	+4.6%	0.4487	No	Neg. Binomial
Arterial	2025 vs 2024	0.510	0.459	0.901	[0.829, 0.978]	-9.9%	0.0131 *	Yes	Poisson
Urban Collectors	2025 vs 2024	0.775	0.760	1.012	[0.918, 1.115]	+1.2%	0.8160	No	Neg. Binomial
Local Access	2025 vs 2024	1.021	0.999	1.003	[0.899, 1.118]	+0.3%	0.9628	No	Neg. Binomial

FSD (Supervised, HW4)

Road Class	Comparison	2024 Rate	2025 Rate	IRR	95% CI	Effect Size	p-value	Significant	Model Used
Highway	2025 vs 2024	0.101	0.118	1.171	[0.785, 1.746]	+17.1%	0.4390	No	Poisson
Arterial	2025 vs 2024	0.241	0.144	0.597	[0.291, 1.226]	-40.3%	0.1600	No	Poisson
Urban Collectors	2025 vs 2024	0.672	0.383	0.571	[0.418, 0.778]	-42.9%	0.0004 ***	Yes	Poisson
Local Access	2025 vs 2024	1.039	0.597	0.574	[0.317, 1.040]	-42.6%	0.0673 †	Approaching	Poisson

Part 3: Major Incidents - Quarterly Trends (Q1 2024 to Q4 2025, North America)

Quarterly Trends: Each chart shows the trend in major collision rates over time (Q1 2024 to Q4 2025) for one road class. Error bars represent 95% confidence intervals. Data includes only January to October for both years.

Note: Quarterly rates are subject to seasonal variations in driving conditions and behavior.

Note on quarterly variation: In early 2024, FSD mileage was lower (e.g., Q1 2024: ~37M total miles across all road classes), producing rate estimates with wide confidence intervals. Individual quarters with small collision counts (1 to 5 events) can produce rate estimates that may appear to overlap with or exceed baseline rates. These do not represent meaningful differences given the limited sample size. The key signals in these charts are the overall trend direction and the progressive narrowing of confidence intervals as FSD mileage has grown over 4x from 2024 to 2025, reflecting increasing statistical precision.

Part 3 Key Takeaways

We observe that FSD (Supervised) shows temporal improvement trends on most road classes.
- Conversely, manually driven Tesla vehicles have not shown similar improvements, with some quarters showing increases in major collision rates.
The key aspects of the quarterly charts are the trend direction and the tightening of confidence intervals over time. In early 2024, FSD mileage was lower (e.g., Q1 2024: 27M highway miles vs. Q3 2025: 400M+ highway miles), resulting in wider confidence intervals. As the fleet has scaled over 4x, rate estimates have stabilized with narrower confidence bounds, reflecting increasing statistical precision.
In early 2024, some non-highway road classes show FSD rates marginally above the manual baseline in individual quarters. These are based on small collision counts (e.g., 1 to 5 events) over limited mileage, producing wide confidence intervals that overlap with the baseline. These do not represent meaningful differences given the insufficient sample size in those periods.

View / Download Data

Quarter	Control Type	Road Class	Total Miles	Events	Rate (per Million Miles)	95% CI Lower	95% CI Upper
2024 Q1	FSD (Supervised, HW4)	Arterial	3,411,176	0	0.00	0.00	1.08
2024 Q1	FSD (Supervised, HW4)	Highway	27,376,549	1	0.04	0.00	0.20
2024 Q1	FSD (Supervised, HW4)	Local Access	873,494	1	1.14	0.03	6.38
2024 Q1	FSD (Supervised, HW4)	Urban Collectors	5,731,544	5	0.87	0.28	2.04
2024 Q1	Manual (No Active Safety Features)	Arterial	6,386,118	20	3.13	1.91	4.84
2024 Q1	Manual (No Active Safety Features)	Highway	22,604,521	10	0.44	0.21	0.81
2024 Q1	Manual (No Active Safety Features)	Local Access	6,122,849	8	1.31	0.56	2.57
2024 Q1	Manual (No Active Safety Features)	Urban Collectors	16,535,055	27	1.63	1.08	2.38
2024 Q1	Manual (With Active Safety Features)	Arterial	586,134,025	347	0.59	0.53	0.66
2024 Q1	Manual (With Active Safety Features)	Highway	2,575,974,646	795	0.31	0.29	0.33
2024 Q1	Manual (With Active Safety Features)	Local Access	617,732,231	671	1.09	1.01	1.17
2024 Q1	Manual (With Active Safety Features)	Urban Collectors	1,854,356,636	1511	0.81	0.77	0.86
2024 Q2	FSD (Supervised, HW4)	Arterial	14,277,759	2	0.14	0.02	0.51
2024 Q2	FSD (Supervised, HW4)	Highway	103,310,832	11	0.11	0.05	0.19
2024 Q2	FSD (Supervised, HW4)	Local Access	4,495,521	4	0.89	0.24	2.28
2024 Q2	FSD (Supervised, HW4)	Urban Collectors	26,904,588	25	0.93	0.60	1.37
2024 Q2	Manual (No Active Safety Features)	Arterial	6,842,888	5	0.73	0.24	1.71
2024 Q2	Manual (No Active Safety Features)	Highway	24,096,126	12	0.50	0.26	0.87
2024 Q2	Manual (No Active Safety Features)	Local Access	6,491,149	5	0.77	0.25	1.80
2024 Q2	Manual (No Active Safety Features)	Urban Collectors	17,472,607	16	0.92	0.52	1.49
2024 Q2	Manual (With Active Safety Features)	Arterial	648,324,396	316	0.49	0.44	0.54
2024 Q2	Manual (With Active Safety Features)	Highway	2,878,754,193	531	0.18	0.17	0.20
2024 Q2	Manual (With Active Safety Features)	Local Access	697,531,663	704	1.01	0.94	1.09
2024 Q2	Manual (With Active Safety Features)	Urban Collectors	2,081,089,348	1560	0.75	0.71	0.79
2024 Q3	FSD (Supervised, HW4)	Arterial	14,243,715	5	0.35	0.11	0.82
2024 Q3	FSD (Supervised, HW4)	Highway	95,294,572	12	0.13	0.07	0.22
2024 Q3	FSD (Supervised, HW4)	Local Access	4,648,292	6	1.29	0.47	2.81
2024 Q3	FSD (Supervised, HW4)	Urban Collectors	26,778,600	13	0.49	0.26	0.83
2024 Q3	Manual (No Active Safety Features)	Arterial	6,479,969	5	0.77	0.25	1.80
2024 Q3	Manual (No Active Safety Features)	Highway	23,311,490	19	0.82	0.49	1.27
2024 Q3	Manual (No Active Safety Features)	Local Access	6,223,617	11	1.77	0.88	3.16
2024 Q3	Manual (No Active Safety Features)	Urban Collectors	16,924,941	23	1.36	0.86	2.04
2024 Q3	Manual (With Active Safety Features)	Arterial	686,058,643	327	0.48	0.43	0.53
2024 Q3	Manual (With Active Safety Features)	Highway	3,089,137,947	537	0.17	0.16	0.19
2024 Q3	Manual (With Active Safety Features)	Local Access	713,506,020	717	1.00	0.93	1.08
2024 Q3	Manual (With Active Safety Features)	Urban Collectors	2,160,574,546	1659	0.77	0.73	0.81
2024 Q4	FSD (Supervised, HW4)	Arterial	9,545,029	3	0.31	0.06	0.92
2024 Q4	FSD (Supervised, HW4)	Highway	62,377,345	5	0.08	0.03	0.19
2024 Q4	FSD (Supervised, HW4)	Local Access	3,457,025	3	0.87	0.18	2.54
2024 Q4	FSD (Supervised, HW4)	Urban Collectors	19,474,494	10	0.51	0.25	0.94
2024 Q4	Manual (No Active Safety Features)	Arterial	2,066,299	1	0.48	0.01	2.70
2024 Q4	Manual (No Active Safety Features)	Highway	7,862,899	7	0.89	0.36	1.83
2024 Q4	Manual (No Active Safety Features)	Local Access	2,183,102	5	2.29	0.74	5.34
2024 Q4	Manual (No Active Safety Features)	Urban Collectors	5,957,603	5	0.84	0.27	1.96
2024 Q4	Manual (With Active Safety Features)	Arterial	238,941,741	111	0.46	0.38	0.56
2024 Q4	Manual (With Active Safety Features)	Highway	1,069,651,028	202	0.19	0.16	0.22
2024 Q4	Manual (With Active Safety Features)	Local Access	269,162,609	255	0.95	0.83	1.07
2024 Q4	Manual (With Active Safety Features)	Urban Collectors	793,511,786	612	0.77	0.71	0.83
2025 Q1	FSD (Supervised, HW4)	Arterial	42,179,728	8	0.19	0.08	0.37
2025 Q1	FSD (Supervised, HW4)	Highway	267,034,528	25	0.09	0.06	0.14
2025 Q1	FSD (Supervised, HW4)	Local Access	16,880,617	4	0.24	0.06	0.61
2025 Q1	FSD (Supervised, HW4)	Urban Collectors	90,668,273	36	0.40	0.28	0.55
2025 Q1	Manual (No Active Safety Features)	Arterial	4,720,896	5	1.06	0.34	2.47
2025 Q1	Manual (No Active Safety Features)	Highway	20,150,073	11	0.55	0.27	0.98
2025 Q1	Manual (No Active Safety Features)	Local Access	5,576,228	5	0.90	0.29	2.09
2025 Q1	Manual (No Active Safety Features)	Urban Collectors	15,685,748	18	1.15	0.68	1.81
2025 Q1	Manual (With Active Safety Features)	Arterial	672,317,374	302	0.45	0.40	0.50
2025 Q1	Manual (With Active Safety Features)	Highway	3,107,496,412	732	0.24	0.22	0.25
2025 Q1	Manual (With Active Safety Features)	Local Access	763,515,310	826	1.08	1.01	1.16
2025 Q1	Manual (With Active Safety Features)	Urban Collectors	2,257,817,954	1798	0.80	0.76	0.83
2025 Q2	FSD (Supervised, HW4)	Arterial	53,979,684	9	0.17	0.08	0.32
2025 Q2	FSD (Supervised, HW4)	Highway	323,055,421	40	0.12	0.09	0.17
2025 Q2	FSD (Supervised, HW4)	Local Access	21,981,171	14	0.64	0.35	1.07
2025 Q2	FSD (Supervised, HW4)	Urban Collectors	112,706,136	50	0.44	0.33	0.58
2025 Q2	Manual (No Active Safety Features)	Arterial	5,132,898	9	1.75	0.80	3.33
2025 Q2	Manual (No Active Safety Features)	Highway	21,634,740	16	0.74	0.42	1.20
2025 Q2	Manual (No Active Safety Features)	Local Access	6,035,960	13	2.15	1.15	3.68
2025 Q2	Manual (No Active Safety Features)	Urban Collectors	16,913,961	31	1.83	1.25	2.60
2025 Q2	Manual (With Active Safety Features)	Arterial	759,715,937	350	0.46	0.41	0.51
2025 Q2	Manual (With Active Safety Features)	Highway	3,469,252,035	675	0.19	0.18	0.21
2025 Q2	Manual (With Active Safety Features)	Local Access	845,614,939	820	0.97	0.90	1.04
2025 Q2	Manual (With Active Safety Features)	Urban Collectors	2,494,722,258	1822	0.73	0.70	0.76
2025 Q3	FSD (Supervised, HW4)	Arterial	75,458,436	11	0.15	0.07	0.26
2025 Q3	FSD (Supervised, HW4)	Highway	443,843,225	51	0.11	0.09	0.15
2025 Q3	FSD (Supervised, HW4)	Local Access	30,267,254	16	0.53	0.30	0.86
2025 Q3	FSD (Supervised, HW4)	Urban Collectors	154,840,393	58	0.37	0.28	0.48
2025 Q3	Manual (No Active Safety Features)	Arterial	5,040,703	7	1.39	0.56	2.86
2025 Q3	Manual (No Active Safety Features)	Highway	21,177,490	12	0.57	0.29	0.99
2025 Q3	Manual (No Active Safety Features)	Local Access	5,824,321	17	2.92	1.70	4.67
2025 Q3	Manual (No Active Safety Features)	Urban Collectors	16,415,285	26	1.58	1.03	2.32
2025 Q3	Manual (With Active Safety Features)	Arterial	801,275,517	359	0.45	0.40	0.50
2025 Q3	Manual (With Active Safety Features)	Highway	3,590,716,519	683	0.19	0.18	0.21
2025 Q3	Manual (With Active Safety Features)	Local Access	880,925,849	833	0.95	0.88	1.01
2025 Q3	Manual (With Active Safety Features)	Urban Collectors	2,581,186,425	1924	0.75	0.71	0.78
2025 Q4	FSD (Supervised, HW4)	Arterial	29,758,213	1	0.03	0.00	0.19
2025 Q4	FSD (Supervised, HW4)	Highway	172,036,605	26	0.15	0.10	0.22
2025 Q4	FSD (Supervised, HW4)	Local Access	12,980,947	15	1.16	0.65	1.91
2025 Q4	FSD (Supervised, HW4)	Urban Collectors	64,374,763	18	0.28	0.17	0.44
2025 Q4	Manual (No Active Safety Features)	Arterial	1,653,263	2	1.21	0.15	4.37
2025 Q4	Manual (No Active Safety Features)	Highway	7,036,774	6	0.85	0.31	1.86
2025 Q4	Manual (No Active Safety Features)	Local Access	1,995,929	5	2.51	0.81	5.85
2025 Q4	Manual (No Active Safety Features)	Urban Collectors	5,607,559	6	1.07	0.39	2.33
2025 Q4	Manual (With Active Safety Features)	Arterial	277,498,512	142	0.51	0.43	0.60
2025 Q4	Manual (With Active Safety Features)	Highway	1,241,540,544	340	0.27	0.25	0.30
2025 Q4	Manual (With Active Safety Features)	Local Access	321,586,154	331	1.03	0.92	1.15
2025 Q4	Manual (With Active Safety Features)	Urban Collectors	924,396,271	731	0.79	0.73	0.85

Part 4: Minor Incidents per Million Miles (2025, North America)

IRR Comparison - Minor Incidents per Million Miles (2025, North America)

Part 4 Key Takeaways

Significant decrease in minor incidents per million miles when using FSD (Supervised) compared to all control types on all non-highway road classes.
- Comparisons to other control types show all 8 point estimates show IRRs below 1, with 6 of the 8 being statistically significant (p < 0.05).
Comparing to manually driven Teslas with active safety features, IRR ranges from 0.409 to 0.521, all with p-values < 0.001.

View / Download Data

Chart Data

Control Type	Road Class	Total Events	Total Miles (M)	Rate per Million Miles	95% CI	Mean	Variance	Var/Mean Ratio	Overdispersed?
Manual (With Active Safety Features)	Highway	9,549	11,409.10	0.837	[0.820, 0.854]	31.411	92.626	2.949	Yes
Manual (With Active Safety Features)	Arterial	3,569	2,510.83	1.421	[1.375, 1.469]	11.740	14.912	1.270	No
Manual (With Active Safety Features)	Urban Collectors	17,670	8,258.19	2.140	[2.108, 2.171]	58.125	113.522	1.953	No
Manual (With Active Safety Features)	Local Access	7,828	2,811.67	2.784	[2.723, 2.846]	25.750	37.383	1.452	No
FSD (Supervised, HW4)	Highway	482	1,205.98	0.400	[0.365, 0.437]	1.586	1.854	1.169	No
FSD (Supervised, HW4)	Arterial	117	201.38	0.581	[0.481, 0.696]	0.385	0.422	1.097	No
FSD (Supervised, HW4)	Urban Collectors	435	422.59	1.029	[0.935, 1.131]	1.431	1.923	1.344	No
FSD (Supervised, HW4)	Local Access	119	82.11	1.449	[1.201, 1.734]	0.391	0.430	1.100	No

Regression Results

Road Class	Comparison	FSD Rate	Baseline Rate	IRR	95% CI	Effect Size	Significant	Model Used
Highway	vs Manual (With Active Safety Features)	0.400	0.837	0.478	[0.433, 0.527]	-52.2%	Yes	Neg. Binomial
Arterial	vs Manual (With Active Safety Features)	0.581	1.421	0.409	[0.340, 0.491]	-59.1%	Yes	Poisson
Urban Collectors	vs Manual (With Active Safety Features)	1.029	2.140	0.481	[0.437, 0.529]	-51.9%	Yes	Poisson
Local Access	vs Manual (With Active Safety Features)	1.449	2.784	0.521	[0.434, 0.624]	-47.9%	Yes	Poisson

Part 5: Minor Incidents - Year over Year Comparison (2024 vs 2025, North America)

Year-over-Year Comparison: Each chart below represents one road class, showing all control types compared between 2024 and 2025. Both years are filtered to the same 10-month period for consistency (January to October).

Note on year-over-year variation: Small fluctuations in rate estimates between years are expected. In 2024, FSD accumulated significantly fewer miles, resulting in rate estimates with wider confidence intervals. Year-over-year changes where confidence intervals overlap or where the IRR is not statistically significant (p > 0.05) should not be interpreted as meaningful trends. The overall regression analysis in Part 4, based on the full 2025 dataset with substantially greater statistical power, provides the most reliable safety comparison.

IRR Comparison - Minor Incidents - Year over Year Comparison (2024 vs 2025, North America)

Part 5 Key Takeaways

FSD (Supervised) has improved in terms of minor collision rates year over year across every road class, with 3 of the 4 being statistically significant (p < 0.05).
- This data, paired with disengagement rate improvements (see study 4) and total mileage increases (see study 1), indicate that FSD (Supervised) is becoming safer, more reliable, and drivers are becoming more comfortable using it.
Manually driven Tesla vehicles have not shown similar improvements.

View / Download Data

Year-over-Year Change Summary

Control Type	Road Class	2024 Events	2024 Miles (M)	2024 Rate	2024 95% CI	2025 Events	2025 Miles (M)	2025 Rate	2025 95% CI	Absolute Change	Percent Change	Mean	Variance	Var/Mean Ratio	Overdispersed?
Manual (With Active Safety Features)	Highway	7,890	9613.52	0.821	[0.803, 0.839]	9,549	11409.01	0.837	[0.820, 0.854]	+0.016	+2.0%	6.282	70.466	11.217	Yes
Manual (With Active Safety Features)	Arterial	3,271	2159.46	1.515	[1.463, 1.568]	3,569	2510.81	1.421	[1.375, 1.469]	-0.093	-6.2%	2.348	10.298	4.386	Yes
Manual (With Active Safety Features)	Urban Collectors	15,117	6889.53	2.194	[2.159, 2.229]	17,670	8258.12	2.140	[2.108, 2.171]	-0.054	-2.5%	11.625	204.519	17.593	Yes
Manual (With Active Safety Features)	Local Access	6,513	2297.93	2.834	[2.766, 2.904]	7,828	2811.64	2.784	[2.723, 2.847]	-0.050	-1.8%	5.150	42.729	8.297	Yes
FSD (Supervised, HW4)	Highway	123	288.36	0.427	[0.355, 0.509]	482	1205.97	0.400	[0.365, 0.437]	-0.027	-6.3%	1.586	1.854	1.169	No
FSD (Supervised, HW4)	Arterial	42	41.48	1.013	[0.730, 1.369]	117	201.38	0.581	[0.481, 0.696]	-0.432	-42.6%	0.385	0.422	1.097	No
FSD (Supervised, HW4)	Urban Collectors	135	78.89	1.711	[1.435, 2.025]	435	422.59	1.029	[0.935, 1.131]	-0.682	-39.8%	1.431	1.923	1.344	No
FSD (Supervised, HW4)	Local Access	34	13.47	2.523	[1.747, 3.526]	119	82.11	1.449	[1.201, 1.734]	-1.074	-42.6%	0.391	0.430	1.100	No

Regression Results

Manual (With Active Safety Features)

Road Class	Comparison	2024 Rate	2025 Rate	IRR	95% CI	Effect Size	p-value	Significant	Model Used
Highway	2025 vs 2024	0.821	0.837	1.082	[0.984, 1.191]	+8.2%	0.1046	No	Neg. Binomial
Arterial	2025 vs 2024	1.515	1.421	1.019	[0.915, 1.135]	+1.9%	0.7351	No	Neg. Binomial
Urban Collectors	2025 vs 2024	2.194	2.140	1.031	[0.943, 1.127]	+3.1%	0.5009	No	Neg. Binomial
Local Access	2025 vs 2024	2.834	2.784	1.007	[0.914, 1.110]	+0.7%	0.8875	No	Neg. Binomial

FSD (Supervised, HW4)

Road Class	Comparison	2024 Rate	2025 Rate	IRR	95% CI	Effect Size	p-value	Significant	Model Used
Highway	2025 vs 2024	0.427	0.400	0.937	[0.769, 1.142]	-6.3%	0.5195	No	Poisson
Arterial	2025 vs 2024	1.013	0.581	0.574	[0.403, 0.816]	-42.6%	0.0020 **	Yes	Poisson
Urban Collectors	2025 vs 2024	1.711	1.029	0.602	[0.496, 0.730]	-39.8%	0.0000 ***	Yes	Poisson
Local Access	2025 vs 2024	2.523	1.449	0.574	[0.392, 0.841]	-42.6%	0.0044 **	Yes	Poisson

Part 6: Minor Incidents - Quarterly Trends (Q1 2024 to Q4 2025, North America)

Quarterly Trends: Each chart shows the trend in minor collision rates over time (Q1 2024 to Q4 2025) for one road class. Error bars represent 95% confidence intervals. Data includes only January to October for both years. Manual (No Active Safety Features) is excluded as these vehicles do not log minor collisions.

Note: Quarterly rates are subject to seasonal variations in driving conditions and behavior.

Note on quarterly variation: In early 2024, FSD mileage was substantially lower, producing rate estimates with wide confidence intervals. The key signals in these charts are the overall trend direction and the progressive narrowing of confidence intervals as FSD mileage has grown over 4x from 2024 to 2025, reflecting increasing statistical precision.

Part 6 Key Takeaways

We observe that FSD (Supervised) shows temporal improvement trends on most road classes.
- Conversely, manually driven Tesla vehicles have not shown similar improvements, with some quarters showing increases in minor collision rates.
As with major collisions, the key aspects of the quarterly minor collision charts are the trend direction and confidence interval narrowing. Early 2024 quarters reflect limited FSD exposure with wider rate estimates. The tightening of confidence intervals from 2024 to 2025 reflects the growing statistical reliability of FSD's safety performance data as mileage has increased.

View / Download Data

Quarter	Control Type	Road Class	Total Miles	Events	Rate (per Million Miles)	95% CI Lower	95% CI Upper
2024 Q1	FSD (Supervised, HW4)	Arterial	3,411,176	4	1.17	0.32	3.00
2024 Q1	FSD (Supervised, HW4)	Highway	27,376,549	6	0.22	0.08	0.48
2024 Q1	FSD (Supervised, HW4)	Local Access	873,494	3	3.43	0.71	10.04
2024 Q1	FSD (Supervised, HW4)	Urban Collectors	5,731,544	5	0.87	0.28	2.04
2024 Q1	Manual (With Active Safety Features)	Arterial	586,134,025	1009	1.72	1.62	1.83
2024 Q1	Manual (With Active Safety Features)	Highway	2,575,974,646	2495	0.97	0.93	1.01
2024 Q1	Manual (With Active Safety Features)	Local Access	617,732,231	1911	3.09	2.96	3.24
2024 Q1	Manual (With Active Safety Features)	Urban Collectors	1,854,356,636	4337	2.34	2.27	2.41
2024 Q2	FSD (Supervised, HW4)	Arterial	14,277,759	17	1.19	0.69	1.91
2024 Q2	FSD (Supervised, HW4)	Highway	103,310,832	42	0.41	0.29	0.55
2024 Q2	FSD (Supervised, HW4)	Local Access	4,495,521	13	2.89	1.54	4.95
2024 Q2	FSD (Supervised, HW4)	Urban Collectors	26,904,588	47	1.75	1.28	2.32
2024 Q2	Manual (With Active Safety Features)	Arterial	648,324,396	922	1.42	1.33	1.52
2024 Q2	Manual (With Active Safety Features)	Highway	2,878,754,193	2241	0.78	0.75	0.81
2024 Q2	Manual (With Active Safety Features)	Local Access	697,531,663	1902	2.73	2.61	2.85
2024 Q2	Manual (With Active Safety Features)	Urban Collectors	2,081,089,348	4479	2.15	2.09	2.22
2024 Q3	FSD (Supervised, HW4)	Arterial	14,243,715	15	1.05	0.59	1.74
2024 Q3	FSD (Supervised, HW4)	Highway	95,294,572	41	0.43	0.31	0.58
2024 Q3	FSD (Supervised, HW4)	Local Access	4,648,292	13	2.80	1.49	4.78
2024 Q3	FSD (Supervised, HW4)	Urban Collectors	26,778,600	52	1.94	1.45	2.55
2024 Q3	Manual (With Active Safety Features)	Arterial	686,058,643	965	1.41	1.32	1.50
2024 Q3	Manual (With Active Safety Features)	Highway	3,089,137,947	2279	0.74	0.71	0.77
2024 Q3	Manual (With Active Safety Features)	Local Access	713,506,020	1985	2.78	2.66	2.91
2024 Q3	Manual (With Active Safety Features)	Urban Collectors	2,160,574,546	4487	2.08	2.02	2.14
2024 Q4	FSD (Supervised, HW4)	Arterial	9,545,029	6	0.63	0.23	1.37
2024 Q4	FSD (Supervised, HW4)	Highway	62,377,345	34	0.55	0.38	0.76
2024 Q4	FSD (Supervised, HW4)	Local Access	3,457,025	5	1.45	0.47	3.38
2024 Q4	FSD (Supervised, HW4)	Urban Collectors	19,474,494	31	1.59	1.08	2.26
2024 Q4	Manual (With Active Safety Features)	Arterial	238,941,741	375	1.57	1.41	1.74
2024 Q4	Manual (With Active Safety Features)	Highway	1,069,651,028	875	0.82	0.76	0.87
2024 Q4	Manual (With Active Safety Features)	Local Access	269,162,609	715	2.66	2.47	2.86
2024 Q4	Manual (With Active Safety Features)	Urban Collectors	793,511,786	1814	2.29	2.18	2.39
2025 Q1	FSD (Supervised, HW4)	Arterial	42,179,728	31	0.73	0.50	1.04
2025 Q1	FSD (Supervised, HW4)	Highway	267,034,528	121	0.45	0.38	0.54
2025 Q1	FSD (Supervised, HW4)	Local Access	16,880,617	22	1.30	0.82	1.97
2025 Q1	FSD (Supervised, HW4)	Urban Collectors	90,668,273	90	0.99	0.80	1.22
2025 Q1	Manual (With Active Safety Features)	Arterial	672,317,374	995	1.48	1.39	1.57
2025 Q1	Manual (With Active Safety Features)	Highway	3,107,496,412	2772	0.89	0.86	0.93
2025 Q1	Manual (With Active Safety Features)	Local Access	763,515,310	2329	3.05	2.93	3.18
2025 Q1	Manual (With Active Safety Features)	Urban Collectors	2,257,817,954	5123	2.27	2.21	2.33
2025 Q2	FSD (Supervised, HW4)	Arterial	53,979,684	33	0.61	0.42	0.86
2025 Q2	FSD (Supervised, HW4)	Highway	323,055,421	116	0.36	0.30	0.43
2025 Q2	FSD (Supervised, HW4)	Local Access	21,981,171	35	1.59	1.11	2.21
2025 Q2	FSD (Supervised, HW4)	Urban Collectors	112,706,136	102	0.91	0.74	1.10
2025 Q2	Manual (With Active Safety Features)	Arterial	759,715,937	1039	1.37	1.29	1.45
2025 Q2	Manual (With Active Safety Features)	Highway	3,469,252,035	2813	0.81	0.78	0.84
2025 Q2	Manual (With Active Safety Features)	Local Access	845,614,939	2289	2.71	2.60	2.82
2025 Q2	Manual (With Active Safety Features)	Urban Collectors	2,494,722,258	5177	2.08	2.02	2.13
2025 Q3	FSD (Supervised, HW4)	Arterial	75,458,436	37	0.49	0.35	0.68
2025 Q3	FSD (Supervised, HW4)	Highway	443,843,225	170	0.38	0.33	0.45
2025 Q3	FSD (Supervised, HW4)	Local Access	30,267,254	45	1.49	1.08	1.99
2025 Q3	FSD (Supervised, HW4)	Urban Collectors	154,840,393	176	1.14	0.97	1.32
2025 Q3	Manual (With Active Safety Features)	Arterial	801,275,517	1088	1.36	1.28	1.44
2025 Q3	Manual (With Active Safety Features)	Highway	3,590,716,519	2836	0.79	0.76	0.82
2025 Q3	Manual (With Active Safety Features)	Local Access	880,925,849	2308	2.62	2.51	2.73
2025 Q3	Manual (With Active Safety Features)	Urban Collectors	2,581,186,425	5266	2.04	1.99	2.10
2025 Q4	FSD (Supervised, HW4)	Arterial	29,758,213	16	0.54	0.31	0.87
2025 Q4	FSD (Supervised, HW4)	Highway	172,036,605	75	0.44	0.34	0.55
2025 Q4	FSD (Supervised, HW4)	Local Access	12,980,947	17	1.31	0.76	2.10
2025 Q4	FSD (Supervised, HW4)	Urban Collectors	64,374,763	67	1.04	0.81	1.32
2025 Q4	Manual (With Active Safety Features)	Arterial	277,498,512	447	1.61	1.46	1.77
2025 Q4	Manual (With Active Safety Features)	Highway	1,241,540,544	1128	0.91	0.86	0.96
2025 Q4	Manual (With Active Safety Features)	Local Access	321,586,154	902	2.80	2.62	2.99
2025 Q4	Manual (With Active Safety Features)	Urban Collectors	924,396,271	2104	2.28	2.18	2.38

Part 7: Regional Comparison - Major Collisions (EU vs NA, 2025)

Regional Comparison - Major Collisions: This analysis compares collision rates between Europe and North America for customer fleet only (non-engineering). Each chart shows one road class with grouped bars comparing EU (green) and NA (blue) rates side-by-side for each control type. Error bars represent 95% confidence intervals. The forest plot shows Incidence Rate Ratios (IRR) where IRR < 1 indicates Europe has a lower rate (better) and IRR > 1 indicates Europe has a higher rate (worse). Data from 2025 (January to October).

Note on cross-regional applicability: FSD collision data in this comparison is from North America. For direct EU vs NA FSD performance comparisons, Study 2: Surrogate Safety Metrics presents AEB activation rates, acceleration events, horn usage, and blinker usage collected from both North American customer fleet and European engineering fleet operations, showing consistent or improved FSD performance on European roads. Study 6 and Study 7 provide additional European FSD validation data.

Part 7 Key Takeaways

European driving is statistically safer than North American driving across all control types and most road classes for major collisions.
- Despite European roads being statistically safer, North American FSD (Supervised) shows <1 IRR for 7 of 8 comparisons, with 6 of 8 statistically significant (p < 0.05), indicating that FSD (Supervised) in North America achieves equivalent or better safety performance relative to European manual driving baselines.
While this comparison uses North American FSD collision data, direct FSD performance comparisons between EU and NA are available in Study 2: Surrogate Safety Metrics, which shows FSD surrogate safety metrics (AEB activation rates, acceleration events, horn usage, and blinker usage) collected from both North American customer fleet and European engineering fleet operations. These metrics demonstrate consistent or improved FSD performance on European roads relative to North American roads.

View / Download Data

Region	Control Type	Road Class	Major Collisions	Total Miles	Rate (per Million Miles)	95% CI
Europe	Manual (No Active Safety Features)	Arterial	5	7,809,535	0.640	[0.208, 1.494]
Europe	Manual (No Active Safety Features)	Highway	13	18,783,971	0.692	[0.369, 1.183]
Europe	Manual (No Active Safety Features)	Local Access	6	4,417,711	1.358	[0.498, 2.956]
Europe	Manual (No Active Safety Features)	Urban Collectors	10	11,789,601	0.848	[0.407, 1.560]
Europe	Manual (With Active Safety Features)	Arterial	340	1,348,783,243	0.252	[0.226, 0.280]
Europe	Manual (With Active Safety Features)	Highway	414	3,523,620,535	0.117	[0.106, 0.129]
Europe	Manual (With Active Safety Features)	Local Access	503	812,741,338	0.619	[0.566, 0.675]
Europe	Manual (With Active Safety Features)	Urban Collectors	972	2,141,127,296	0.454	[0.426, 0.483]
North America	FSD (Supervised, HW4)	Arterial	29	201,377,865	0.144	[0.096, 0.207]
North America	FSD (Supervised, HW4)	Highway	142	1,205,981,745	0.118	[0.099, 0.139]
North America	FSD (Supervised, HW4)	Local Access	49	82,110,744	0.597	[0.441, 0.789]
North America	FSD (Supervised, HW4)	Urban Collectors	162	422,593,313	0.383	[0.327, 0.447]
North America	Manual (No Active Safety Features)	Arterial	23	16,547,882	1.390	[0.881, 2.086]
North America	Manual (No Active Safety Features)	Highway	45	69,999,597	0.643	[0.469, 0.860]
North America	Manual (No Active Safety Features)	Local Access	40	19,432,632	2.058	[1.471, 2.803]
North America	Manual (No Active Safety Features)	Urban Collectors	81	54,623,024	1.483	[1.178, 1.843]
North America	Manual (With Active Safety Features)	Arterial	1153	2,510,828,318	0.459	[0.433, 0.487]
North America	Manual (With Active Safety Features)	Highway	2430	11,409,102,069	0.213	[0.205, 0.222]
North America	Manual (With Active Safety Features)	Local Access	2810	2,811,665,301	0.999	[0.963, 1.037]
North America	Manual (With Active Safety Features)	Urban Collectors	6275	8,258,191,430	0.760	[0.741, 0.779]

Regression Results

Control Type	Road Class	EU Count	EU Rate	NA Count	NA Rate	IRR (EU/NA)	95% CI Lower	95% CI Upper	p-value	Significance	% Difference	Direction
Manual (No Active Safety Features)	Arterial	5	0.640	23	1.390	0.461	0.175	1.212	0.1162	ns	53.9%	EU lower (better)
Manual (No Active Safety Features)	Highway	13	0.692	45	0.643	1.077	0.581	1.996	0.8148	ns	7.7%	EU higher (worse)
Manual (No Active Safety Features)	Local Access	6	1.358	40	2.058	0.660	0.280	1.556	0.3423	ns	34.0%	EU lower (better)
Manual (No Active Safety Features)	Urban Collectors	10	0.848	81	1.483	0.572	0.297	1.103	0.0956	†	42.8%	EU lower (better)
Manual (With Active Safety Features)	Arterial	340	0.252	1153	0.459	0.549	0.486	0.620	<0.0001	***	45.1%	EU lower (better)
Manual (With Active Safety Features)	Highway	414	0.117	2430	0.213	0.552	0.497	0.612	<0.0001	***	44.8%	EU lower (better)
Manual (With Active Safety Features)	Local Access	503	0.619	2810	0.999	0.619	0.563	0.681	<0.0001	***	38.1%	EU lower (better)
Manual (With Active Safety Features)	Urban Collectors	972	0.454	6275	0.760	0.597	0.558	0.639	<0.0001	***	40.3%	EU lower (better)

EU vs NA FSD Comparison Results

EU Control Type	Road Class	EU Rate	NA FSD Rate	IRR (NA FSD/EU)	95% CI Lower	95% CI Upper	p-value	Significance	% Difference	Direction	label	road_sort	control_sort
Manual (No Active Safety Features)	Local Access	1.358	0.597	0.439	0.188	1.026	0.0573	†	56.1%	FSD lower (better)	EU Manual (No Active Safety) - Local Access	3	0
Manual (No Active Safety Features)	Urban Collectors	0.848	0.383	0.452	0.239	0.856	0.0148	*	54.8%	FSD lower (better)	EU Manual (No Active Safety) - Urban Collectors	2	0
Manual (No Active Safety Features)	Arterial	0.640	0.144	0.225	0.087	0.581	0.0021	**	77.5%	FSD lower (better)	EU Manual (No Active Safety) - Arterial	1	0
Manual (No Active Safety Features)	Highway	0.692	0.118	0.170	0.096	0.300	<0.0001	***	83.0%	FSD lower (better)	EU Manual (No Active Safety) - Highway	0	0
Manual (With Active Safety Features)	Local Access	0.619	0.597	0.964	0.719	1.293	0.8077	ns	3.6%	FSD lower (better)	EU Manual (Active Safety) - Local Access	3	1
Manual (With Active Safety Features)	Urban Collectors	0.454	0.383	0.844	0.715	0.997	0.0463	*	15.6%	FSD lower (better)	EU Manual (Active Safety) - Urban Collectors	2	1
Manual (With Active Safety Features)	Arterial	0.252	0.144	0.571	0.391	0.835	0.0038	**	42.9%	FSD lower (better)	EU Manual (Active Safety) - Arterial	1	1
Manual (With Active Safety Features)	Highway	0.117	0.118	1.002	0.828	1.213	0.9823	ns	0.2%	EU lower (better)	EU Manual (Active Safety) - Highway	0	1

Part 8: Regional Comparison - Minor Collisions (EU vs NA, 2025)

Regional Comparison - Minor Collisions: This analysis compares minor collision rates between Europe and North America for the active customer fleet. Each chart shows one road class with grouped bars comparing EU (green) and NA (blue) rates side-by-side for each control type. Error bars represent 95% confidence intervals. The forest plot shows Incidence Rate Ratios (IRR) where IRR < 1 indicates Europe has a lower rate (better) and IRR > 1 indicates Europe has a higher rate (worse). Data from 2025 (January to October).

Note on cross-regional applicability: FSD collision data in this comparison is from North America. For direct EU vs NA FSD performance comparisons, see Study 2 (surrogate safety metrics from both regions), Study 6 (European engineering fleet testing), and Study 7 (European fixed routes testing).

Part 8 Key Takeaways

European driving is statistically safer than North American driving across all control types and most road classes for minor collisions.
- Despite European roads being statistically safer, North American FSD (Supervised) shows statistically significant decreases in minor collisions for all comparisons, indicating equivalent or better safety performance relative to European manual driving baselines.
For direct EU vs NA FSD performance comparisons on surrogate safety metrics, see Study 2. Additionally, Study 6: Engineering Fleet Testing shows over 793K FSD miles driven across 8 European countries with zero collisions due to FSD performance, and Study 7: Fixed Routes Testing shows zero safety-critical events across 230,000+ scenario tests in 6 European cities.

View / Download Data

Region	Control Type	Road Class	Minor Collisions	Total Miles	Rate (per Million Miles)	95% CI
Europe	Manual (No Active Safety Features)	Arterial	0	7,809,535	0.000	[0.000, 0.472]
Europe	Manual (No Active Safety Features)	Highway	0	18,783,971	0.000	[0.000, 0.196]
Europe	Manual (No Active Safety Features)	Local Access	0	4,417,711	0.000	[0.000, 0.835]
Europe	Manual (No Active Safety Features)	Urban Collectors	0	11,789,601	0.000	[0.000, 0.313]
Europe	Manual (With Active Safety Features)	Arterial	1105	1,348,783,243	0.819	[0.772, 0.869]
Europe	Manual (With Active Safety Features)	Highway	1763	3,523,620,535	0.500	[0.477, 0.524]
Europe	Manual (With Active Safety Features)	Local Access	1543	812,741,338	1.899	[1.805, 1.996]
Europe	Manual (With Active Safety Features)	Urban Collectors	2531	2,141,127,296	1.182	[1.136, 1.229]
North America	FSD (Supervised, HW4)	Arterial	117	201,377,865	0.581	[0.481, 0.696]
North America	FSD (Supervised, HW4)	Highway	482	1,205,981,745	0.400	[0.365, 0.437]
North America	FSD (Supervised, HW4)	Local Access	119	82,110,744	1.449	[1.201, 1.734]
North America	FSD (Supervised, HW4)	Urban Collectors	435	422,593,313	1.029	[0.935, 1.131]
North America	Manual (No Active Safety Features)	Arterial	0	16,547,882	0.000	[0.000, 0.223]
North America	Manual (No Active Safety Features)	Highway	0	69,999,597	0.000	[0.000, 0.053]
North America	Manual (No Active Safety Features)	Local Access	0	19,432,632	0.000	[0.000, 0.190]
North America	Manual (No Active Safety Features)	Urban Collectors	0	54,623,024	0.000	[0.000, 0.068]
North America	Manual (With Active Safety Features)	Arterial	3569	2,510,828,318	1.421	[1.375, 1.469]
North America	Manual (With Active Safety Features)	Highway	9549	11,409,102,069	0.837	[0.820, 0.854]
North America	Manual (With Active Safety Features)	Local Access	7828	2,811,665,301	2.784	[2.723, 2.846]
North America	Manual (With Active Safety Features)	Urban Collectors	17670	8,258,191,430	2.140	[2.108, 2.171]

Regression Results

Control Type	Road Class	EU Count	EU Rate	NA Count	NA Rate	IRR (EU/NA)	95% CI Lower	95% CI Upper	p-value	Significance	% Difference	Direction
Manual (With Active Safety Features)	Arterial	1105	0.819	3569	1.421	0.576	0.539	0.617	<0.0001	***	42.4%	EU lower (better)
Manual (With Active Safety Features)	Highway	1763	0.500	9549	0.837	0.598	0.568	0.629	<0.0001	***	40.2%	EU lower (better)
Manual (With Active Safety Features)	Local Access	1543	1.899	7828	2.784	0.682	0.646	0.720	<0.0001	***	31.8%	EU lower (better)
Manual (With Active Safety Features)	Urban Collectors	2531	1.182	17670	2.140	0.552	0.530	0.576	<0.0001	***	44.8%	EU lower (better)

EU vs NA FSD Comparison Results

EU Control Type	Road Class	EU Rate	NA FSD Rate	IRR (NA FSD/EU)	95% CI Lower	95% CI Upper	p-value	Significance	% Difference	Direction	label	road_sort	control_sort
Manual (With Active Safety Features)	Local Access	1.899	1.449	0.763	0.634	0.920	0.0045	**	23.7%	FSD lower (better)	EU Manual (Active Safety) - Local Access	3	1
Manual (With Active Safety Features)	Urban Collectors	1.182	1.029	0.871	0.787	0.964	0.0077	**	12.9%	FSD lower (better)	EU Manual (Active Safety) - Urban Collectors	2	1
Manual (With Active Safety Features)	Arterial	0.819	0.581	0.709	0.586	0.858	0.0004	***	29.1%	FSD lower (better)	EU Manual (Active Safety) - Arterial	1	1
Manual (With Active Safety Features)	Highway	0.500	0.400	0.799	0.722	0.883	<0.0001	***	20.1%	FSD lower (better)	EU Manual (Active Safety) - Highway	0	1

Study 6: Engineering Fleet Testing

The data collected from this program is to provide foundational evidence of system capability, robustness, and safety under real-world European conditions. The program includes:

Golden Manual Data Collection: Trained operators manually drive the vehicle to collect high-quality, labeled training data for the end-to-end model. This data is used to improve the system's performance and is distinct from the FSD (Supervised) validation data.
FSD (Supervised) Validation: Extensive testing of the FSD (Supervised) system under real-world conditions through mileage collection.

For additional statistics from the engineering validation program, please reference the fixed routes and surrogate safety metrics sections

For video clip evidence of competency, please reference the video clips section of this dashboard.

Exemption Relevance: This study contributes to the overall safety case underlying all exemption requests, demonstrating over 793K FSD miles across 8 European countries with zero major or minor collisions observed due to FSD performance.

Study Takeaways

FSD (Supervised) has driven roughly 1.0 million miles (1.6 million kilometers) in Europe with zero major or minor collisions observed due to FSD performance, demonstrating FSD's ability to operate safely in European conditions under rigorous safety standards. For surrogate safety metrics from the engineering fleet, please reference the surrogate safety metrics study.
- Tesla has done it's dilligence to make sure that the largest and most populous European countries have been included in the engineering fleet testing program with significant splits of mileage.
Golden Manual data collection has accumulated roughly 2.0 million miles (3.1 million kilometers) to support model training and validation.

Mileage Accumulated in European Countries

All European Countries Tested

Golden Manual Data Collection

1,953,566 miles

FSD (Supervised) Validation

993,940 miles

FSD (Supervised)

40,064 miles

View / Download Data

Mileage Data and Conservative Upper Bounds

Control Type	Road Class	Total Miles (M)
Manual (With Active Safety Features)	Highway	0.94
Manual (With Active Safety Features)	Arterial	0.31
Manual (With Active Safety Features)	Urban Collectors	0.45
Manual (With Active Safety Features)	Local Access	0.25
FSD (Supervised, HW4)	Highway	0.55
FSD (Supervised, HW4)	Arterial	0.15
FSD (Supervised, HW4)	Urban Collectors	0.24
FSD (Supervised, HW4)	Local Access	0.06

Study 7: Fixed Routes Testing

The preceding studies establish FSD (Supervised) safety performance through large-scale statistical analysis of real-world driving data. Fixed route testing complements this evidence with structured behavioral validation, demonstrating that the system correctly handles specific, predefined scenario types in European driving conditions.

Fixed routes are designed to concentrate challenging driving interactions into short, repeatable paths (29-55 km per route). A single route execution exposes the system to hundreds of complex scenarios, including roundabouts, cyclist yields, tram interactions, and dense urban intersections, providing a density of edge cases that would require significantly more open-road driving to encounter naturally. Critically, FSD (Supervised) has not been trained on any of these specific routes, so the results reflect the system's ability to generalize its learned driving behavior to novel European road environments.

Testing was conducted across 6 major European cities (Amsterdam, Barcelona, Rome, Paris, Munich, Copenhagen), each representing distinct traffic cultures and infrastructure types. Data covers the period from June 1, 2025 to November 27, 2025. Results are presented below.

Exemption Relevance: This study contributes to the overall safety case underlying all exemption requests, with zero safety-critical events across over 230,000 scenario tests, where each test is a unique interaction with a predefined road element such as a stop sign, traffic light, roundabout, or crosswalk. It provides supporting evidence for Exemption Request 4 (lateral acceleration limits) through a 98.9% pass rate across 6,974 roundabout tests, and for Exemption Requests 1, 2, and 3 (SIMs and HOR withholding) through safe completion of complex system-initiated maneuvers including lane changes, roundabouts, and intersection navigation.

Key Takeaways

Across all fixed route testing, there were zero safety-critical events observed, demonstrating FSD's ability to operate safely in European road conditions through complex maneuvers including challenging interactions with pedestrians, cyclists, and dynamic traffic scenarios. This testing was conducted by certified drivers trained by Tesla.
FSD (Supervised) completed over 230,000 scenario tests across 6 major European cities, covering a wide range of driving scenarios, traffic laws, and interaction types. Each scenario test is a unique interaction with a predefined road element on the route (e.g., navigating a roundabout, yielding at a crosswalk, responding to a traffic light, or passing through an intersection). A test is scored as a "pass" if FSD handles the interaction safely without driver disengagement, or a "fail" if the safety driver intervenes.
Overall, FSD (Supervised) achieved a pass rate of at least ninety two percent (92.0%) across all interaction types and environmental conditions, indicating robust performance in navigating complex urban environments.
Stratification by time of day and weather show strong performance in varying environmental conditions.
- The user manual includes system limitations for varying conditions as they could potentially disturb the visibility of the cameras. The exact text in the user manual is "Visibility is critical for Full Self-Driving (Supervised) to operate. Low visibility, such as low light or poor weather conditions (rain, snow, direct sun, fog, etc.) can significantly degrade performance."

Fixed Route Overview

Methodology

Fixed routes are designed using Tesla's mapping capabilities to identify locations containing high concentrations of challenging scenarios. Routes range from 29 km to 55 km in length and intentionally concentrate difficult infrastructure variations and dynamic interactions far beyond typical driving experiences. This creates a controlled environment to validate the system's ability to handle edge cases and complex scenarios.

Each route exposes the system to hundreds of challenging interactions per run, including multi-lane roundabouts with complex right-of-way rules, dense cyclist environments requiring constant prediction and yielding, dynamic pedestrian crossings, tram interactions, and narrow dike roads with oncoming traffic negotiation. These scenarios test the system's generalization capabilities without any route-specific training or fine-tuning.

Each route is driven repeatedly with a safety driver present to disengage when necessary. Only disengagements within pre-defined test scenarios are counted to ensure comparability across test runs.

Test Cities and Routes

The following fixed routes were designed and tested across 6 major European cities, each presenting unique infrastructure challenges and traffic patterns characteristic of European urban environments.

Amsterdam

Distance: 56.8 km

Duration: 1h 32m

Executions: 99

Munich

Distance: 34.3 km

Duration: 59m

Executions: 57

Barcelona

Distance: 34.3 km

Duration: 1h 15m

Executions: 184

Rome

Distance: 30.5 km

Duration: 1h 11m

Executions: 134

Paris

Distance: 38.4 km

Duration: 1h 1m

Executions: 105

Copenhagen

Distance: 27.6 km

Duration: 55m

Executions: 32

Test Criteria

Pass: No safety driver disengagement required; scenario handled safely and in compliance with traffic laws.
Fail: Safety driver disengagement deemed necessary to maintain safe operations or prevent traffic law violations.

Tested Scenarios

Scenario	Description
Intersections	Navigate signalized and unsignalized intersections while managing vehicle interactions
Turns (Left/Right)	Execute turns at intersections with signal adherence, trajectory optimization, and yielding to crosswalks/cyclists
Crosswalks	Detect and yield to pedestrian crossings, ensuring timely stops and safe resumption
Traffic Lights	Detect and respond to traffic signals, including complex multi-direction signals
Roundabouts	Navigate roundabouts, manage right-of-way, handle multi-lane merging and blinker usage
Yield Scenarios	Prioritize pedestrians and cyclists at sidewalks and bike lanes
Cyclist Roads	Interact with high volumes of cyclists; detect, predict movement, and safely overtake or yield
Tram Lanes	Recognize and respect tram right-of-way; avoid interference with tram operations
Bus Lanes	Comply with bus-only lane restrictions; detect markings and avoid unauthorized entry
Bus Stops	Detect stopped buses and yield to buses leaving stops in built-up areas
Stop Signs	Detect stop signs and perform complete stops with proper yielding to traffic
School Zones	Adhere to reduced speed limits with heightened awareness of pedestrian activity
Living Streets	Navigate low-speed, pedestrian-priority zones where vehicles yield to non-motorized users
Road Narrowing	Adjust speed and trajectory in constrained spaces
Bridges	Maintain stability on elevated structures with environmental factors like wind or narrow lanes
Dike Roads	Navigate narrow, elevated roads with limited visibility while negotiating with other traffic

Note: Fixed routes provide concentrated exposure to challenging scenarios for validation purposes. They are not intended to be statistically representative of the overall operational design domain, but rather to stress-test the system's learned behaviors and edge case handling without prior training on these specific routes.

Each test run is classified as either a "pass" (successful navigation without disengagement) or "fail" (requiring driver disengagement). Results are aggregated by interaction type and stratified by environmental conditions (weather and time of day). Pass rate confidence intervals are calculated using the exact binomial method (Clopper-Pearson intervals), providing conservative 95% confidence bounds.

Fixed Route Test Results

Filter by City:

Filter by Component:

Summary Results (All Cities, All Components)

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI	Safety-Critical Pass Rate ⓘ
Intersection	72,658	572	73,230	99.22%	[99.15%, 99.28%]	100%
Crosswalk	47,313	161	47,474	99.66%	[99.60%, 99.71%]	100%
Traffic Light	33,178	846	34,024	97.51%	[97.34%, 97.67%]	100%
Yield Road	20,263	83	20,346	99.59%	[99.49%, 99.67%]	100%
Bus Stop	18,741	8	18,749	99.96%	[99.92%, 99.98%]	100%
Yield Bikelane	7,098	10	7,108	99.86%	[99.74%, 99.92%]	100%
Roundabout	6,900	74	6,974	98.94%	[98.67%, 99.15%]	100%
Yield Sidewalk	4,197	5	4,202	99.88%	[99.72%, 99.95%]	100%
Digital Speed Sign	3,471	68	3,539	98.08%	[97.57%, 98.48%]	100%
Priority	2,716	4	2,720	99.85%	[99.62%, 99.94%]	100%
Stop Sign	2,516	191	2,707	92.94%	[91.92%, 93.85%]	100%
Road Narrowing	2,262	20	2,282	99.12%	[98.65%, 99.43%]	100%
Bridge	1,286	12	1,298	99.08%	[98.39%, 99.47%]	100%
Avoid Bus Lane	1,234	95	1,329	92.85%	[91.34%, 94.12%]	100%
Yield Tram	1,181	0	1,181	100.00%	[99.68%, 100.00%]	100%
Yield Bus Lane	1,047	7	1,054	99.34%	[98.64%, 99.68%]	100%
Digital sign	894	41	935	95.61%	[94.11%, 96.75%]	100%
Highway Entry	827	2	829	99.76%	[99.12%, 99.93%]	100%
Priority Sign	814	6	820	99.27%	[98.41%, 99.66%]	100%
Highway Exit	810	51	861	94.08%	[92.30%, 95.47%]	100%
School Zone	649	3	652	99.54%	[98.66%, 99.84%]	100%
Dynamic Barrier	336	3	339	99.12%	[97.43%, 99.70%]	100%
Living Street	196	3	199	98.49%	[95.66%, 99.49%]	100%
U Turn	167	2	169	98.82%	[95.79%, 99.67%]	100%
Intersection	18,777	221	18,998	98.84%	[98.67%, 98.98%]	100%
Crosswalk	7,354	20	7,374	99.73%	[99.58%, 99.82%]	100%
Traffic Light	6,582	300	6,882	95.64%	[95.13%, 96.10%]	100%
Yield Bikelane	5,316	9	5,325	99.83%	[99.68%, 99.91%]	100%
Yield Road	3,673	9	3,682	99.76%	[99.54%, 99.87%]	100%
Digital Speed Sign	2,703	1	2,704	99.96%	[99.79%, 99.99%]	100%
Roundabout	1,641	5	1,646	99.70%	[99.29%, 99.87%]	100%
Yield Sidewalk	1,522	0	1,522	100.00%	[99.75%, 100.00%]	100%
Bus Stop	1,513	3	1,516	99.80%	[99.42%, 99.93%]	100%
Road Narrowing	1,500	13	1,513	99.14%	[98.54%, 99.50%]	100%
Yield Tram	1,181	0	1,181	100.00%	[99.68%, 100.00%]	100%
Yield Bus Lane	843	1	844	99.88%	[99.33%, 99.98%]	100%
Bridge	831	12	843	98.58%	[97.53%, 99.18%]	100%
Avoid Bus Lane	504	1	505	99.80%	[98.89%, 99.97%]	100%
Priority Sign	499	6	505	98.81%	[97.43%, 99.45%]	100%
School Zone	337	0	337	100.00%	[98.87%, 100.00%]	100%
Dynamic Barrier	336	3	339	99.12%	[97.43%, 99.70%]	100%
Living Street	169	2	171	98.83%	[95.84%, 99.68%]	100%
Highway Entry	169	0	169	100.00%	[97.78%, 100.00%]	100%
Highway Exit	168	1	169	99.41%	[96.72%, 99.90%]	100%
U Turn	167	2	169	98.82%	[95.79%, 99.67%]	100%
Intersection	17,351	261	17,612	98.52%	[98.33%, 98.69%]	100%
Crosswalk	14,228	89	14,317	99.38%	[99.24%, 99.49%]	100%
Traffic Light	11,439	361	11,800	96.94%	[96.61%, 97.24%]	100%
Bus Stop	4,636	2	4,638	99.96%	[99.84%, 99.99%]	100%
Yield Road	2,469	45	2,514	98.21%	[97.61%, 98.66%]	100%
Roundabout	1,107	54	1,161	95.35%	[93.98%, 96.42%]	100%
Stop Sign	708	64	772	91.71%	[89.55%, 93.45%]	100%
Digital Speed Sign	545	34	579	94.13%	[91.91%, 95.77%]	100%
Avoid Bus Lane	194	0	194	100.00%	[98.06%, 100.00%]	100%
Yield Bikelane	193	0	193	100.00%	[98.05%, 100.00%]	100%
Highway Entry	191	2	193	98.96%	[96.30%, 99.72%]	100%
Highway Exit	144	49	193	74.61%	[68.03%, 80.23%]	100%
Intersection	1,658	6	1,664	99.64%	[99.22%, 99.83%]	100%
Bus Stop	1,280	0	1,280	100.00%	[99.70%, 100.00%]	100%
Traffic Light	1,122	30	1,152	97.40%	[96.31%, 98.17%]	100%
Crosswalk	1,053	3	1,056	99.72%	[99.17%, 99.90%]	100%
Yield Bikelane	959	1	960	99.90%	[99.41%, 99.98%]	100%
Yield Road	640	0	640	100.00%	[99.40%, 100.00%]	100%
Digital Speed Sign	223	33	256	87.11%	[82.45%, 90.67%]	100%
Roundabout	190	2	192	98.96%	[96.28%, 99.71%]	100%
Yield Sidewalk	128	0	128	100.00%	[97.09%, 100.00%]	100%
Road Narrowing	93	1	94	98.94%	[94.22%, 99.81%]	100%
Highway Exit	64	0	64	100.00%	[94.34%, 100.00%]	100%
Highway Entry	32	0	32	100.00%	[89.28%, 100.00%]	100%
Intersection	11,007	42	11,049	99.62%	[99.49%, 99.72%]	100%
Traffic Light	3,332	43	3,375	98.73%	[98.29%, 99.05%]	100%
Priority	2,716	4	2,720	99.85%	[99.62%, 99.94%]	100%
Yield Sidewalk	2,235	2	2,237	99.91%	[99.67%, 99.98%]	100%
Bus Stop	1,444	1	1,445	99.93%	[99.61%, 99.99%]	100%
Digital sign	894	41	935	95.61%	[94.11%, 96.75%]	100%
Crosswalk	679	1	680	99.85%	[99.17%, 99.97%]	100%
Roundabout	510	0	510	100.00%	[99.25%, 100.00%]	100%
Road Narrowing	252	3	255	98.82%	[96.60%, 99.60%]	100%
Stop Sign	147	3	150	98.00%	[94.29%, 99.32%]	100%
Highway Entry	85	0	85	100.00%	[95.68%, 100.00%]	100%
Highway Exit	85	0	85	100.00%	[95.68%, 100.00%]	100%
Living Street	27	1	28	96.43%	[82.29%, 99.37%]	100%
Crosswalk	13,514	31	13,545	99.77%	[99.68%, 99.84%]	100%
Intersection	5,544	21	5,565	99.62%	[99.42%, 99.75%]	100%
Yield Road	4,394	16	4,410	99.64%	[99.41%, 99.78%]	100%
Traffic Light	4,026	69	4,095	98.32%	[97.87%, 98.67%]	100%
Roundabout	3,032	13	3,045	99.57%	[99.27%, 99.75%]	100%
Bus Stop	2,728	2	2,730	99.93%	[99.73%, 99.98%]	100%
Stop Sign	931	14	945	98.52%	[97.53%, 99.12%]	100%
Yield Bikelane	630	0	630	100.00%	[99.39%, 100.00%]	100%
Road Narrowing	417	3	420	99.29%	[97.92%, 99.76%]	100%
Priority Sign	315	0	315	100.00%	[98.80%, 100.00%]	100%
Bridge	315	0	315	100.00%	[98.80%, 100.00%]	100%
Yield Sidewalk	312	3	315	99.05%	[97.24%, 99.68%]	100%
School Zone	312	3	315	99.05%	[97.24%, 99.68%]	100%
Highway Entry	210	0	210	100.00%	[98.20%, 100.00%]	100%
Highway Exit	209	1	210	99.52%	[97.35%, 99.92%]	100%
Yield Bus Lane	204	6	210	97.14%	[93.91%, 98.68%]	100%
Avoid Bus Lane	204	6	210	97.14%	[93.91%, 98.68%]	100%
Intersection	18,321	21	18,342	99.89%	[99.83%, 99.93%]	100%
Crosswalk	10,485	17	10,502	99.84%	[99.74%, 99.90%]	100%
Yield Road	9,087	13	9,100	99.86%	[99.76%, 99.92%]	100%
Bus Stop	7,140	0	7,140	100.00%	[99.95%, 100.00%]	100%
Traffic Light	6,677	43	6,720	99.36%	[99.14%, 99.52%]	100%
Stop Sign	730	110	840	86.90%	[84.45%, 89.02%]	100%
Roundabout	420	0	420	100.00%	[99.09%, 100.00%]	100%
Avoid Bus Lane	332	88	420	79.05%	[74.90%, 82.67%]	100%
Bridge	140	0	140	100.00%	[97.33%, 100.00%]	100%
Highway Entry	140	0	140	100.00%	[97.33%, 100.00%]	100%
Highway Exit	140	0	140	100.00%	[97.33%, 100.00%]	100%

Performance by Environmental Conditions

Weather Condition

Rain (All Cities, All Components)

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	1,799	9	1,808	99.50%	[99.06%, 99.74%]
Traffic Light	1,028	28	1,056	97.35%	[96.19%, 98.16%]
Crosswalk	704	0	704	100.00%	[99.46%, 100.00%]
Yield Bikelane	512	0	512	100.00%	[99.26%, 100.00%]
Yield Road	352	0	352	100.00%	[98.92%, 100.00%]
Digital Speed Sign	256	0	256	100.00%	[98.52%, 100.00%]
Roundabout	160	0	160	100.00%	[97.66%, 100.00%]
Bus Stop	144	0	144	100.00%	[97.40%, 100.00%]
Yield Sidewalk	144	0	144	100.00%	[97.40%, 100.00%]
Road Narrowing	142	2	144	98.61%	[95.08%, 99.62%]
Yield Tram	112	0	112	100.00%	[96.68%, 100.00%]
Yield Bus Lane	80	0	80	100.00%	[95.42%, 100.00%]
Bridge	79	1	80	98.75%	[93.25%, 99.78%]
Avoid Bus Lane	48	0	48	100.00%	[92.59%, 100.00%]
Priority Sign	46	2	48	95.83%	[86.02%, 98.85%]
School Zone	32	0	32	100.00%	[89.28%, 100.00%]
Dynamic Barrier	31	1	32	96.88%	[84.26%, 99.45%]
Highway Entry	16	0	16	100.00%	[80.64%, 100.00%]
Highway Exit	16	0	16	100.00%	[80.64%, 100.00%]
Living Street	16	0	16	100.00%	[80.64%, 100.00%]
U Turn	16	0	16	100.00%	[80.64%, 100.00%]
Intersection	1,075	17	1,092	98.44%	[97.52%, 99.03%]
Crosswalk	884	4	888	99.55%	[98.85%, 99.82%]
Traffic Light	708	24	732	96.72%	[95.17%, 97.79%]
Bus Stop	288	0	288	100.00%	[98.68%, 100.00%]
Yield Road	154	2	156	98.72%	[95.45%, 99.65%]
Roundabout	70	2	72	97.22%	[90.43%, 99.23%]
Stop Sign	44	4	48	91.67%	[80.45%, 96.71%]
Digital Speed Sign	33	3	36	91.67%	[78.17%, 97.13%]
Avoid Bus Lane	12	0	12	100.00%	[75.75%, 100.00%]
Highway Entry	12	0	12	100.00%	[75.75%, 100.00%]
Yield Bikelane	12	0	12	100.00%	[75.75%, 100.00%]
Highway Exit	10	2	12	83.33%	[55.20%, 95.30%]
Intersection	104	0	104	100.00%	[96.44%, 100.00%]
Bus Stop	80	0	80	100.00%	[95.42%, 100.00%]
Traffic Light	70	2	72	97.22%	[90.43%, 99.23%]
Crosswalk	66	0	66	100.00%	[94.50%, 100.00%]
Yield Bikelane	60	0	60	100.00%	[93.98%, 100.00%]
Yield Road	40	0	40	100.00%	[91.24%, 100.00%]
Digital Speed Sign	16	0	16	100.00%	[80.64%, 100.00%]
Roundabout	12	0	12	100.00%	[75.75%, 100.00%]
Yield Sidewalk	8	0	8	100.00%	[67.56%, 100.00%]
Road Narrowing	6	0	6	100.00%	[60.97%, 100.00%]
Highway Exit	4	0	4	100.00%	[51.01%, 100.00%]
Highway Entry	2	0	2	100.00%	[34.24%, 100.00%]
Intersection	1,161	9	1,170	99.23%	[98.54%, 99.59%]
Traffic Light	354	6	360	98.33%	[96.41%, 99.23%]
Priority	288	0	288	100.00%	[98.68%, 100.00%]
Yield Sidewalk	236	0	236	100.00%	[98.40%, 100.00%]
Bus Stop	153	0	153	100.00%	[97.55%, 100.00%]
Digital sign	95	4	99	95.96%	[90.07%, 98.42%]
Crosswalk	72	0	72	100.00%	[94.93%, 100.00%]
Roundabout	54	0	54	100.00%	[93.36%, 100.00%]
Road Narrowing	27	0	27	100.00%	[87.54%, 100.00%]
Stop Sign	16	0	16	100.00%	[80.64%, 100.00%]
Highway Entry	9	0	9	100.00%	[70.09%, 100.00%]
Highway Exit	9	0	9	100.00%	[70.09%, 100.00%]
Living Street	2	0	2	100.00%	[34.24%, 100.00%]
Crosswalk	1,543	5	1,548	99.68%	[99.25%, 99.86%]
Intersection	636	0	636	100.00%	[99.40%, 100.00%]
Yield Road	501	3	504	99.40%	[98.26%, 99.80%]
Traffic Light	464	4	468	99.15%	[97.82%, 99.67%]
Roundabout	345	3	348	99.14%	[97.50%, 99.71%]
Bus Stop	312	0	312	100.00%	[98.78%, 100.00%]
Stop Sign	108	0	108	100.00%	[96.57%, 100.00%]
Yield Bikelane	72	0	72	100.00%	[94.93%, 100.00%]
Road Narrowing	48	0	48	100.00%	[92.59%, 100.00%]
Bridge	36	0	36	100.00%	[90.36%, 100.00%]
Priority Sign	36	0	36	100.00%	[90.36%, 100.00%]
School Zone	36	0	36	100.00%	[90.36%, 100.00%]
Yield Sidewalk	36	0	36	100.00%	[90.36%, 100.00%]
Avoid Bus Lane	24	0	24	100.00%	[86.20%, 100.00%]
Highway Entry	24	0	24	100.00%	[86.20%, 100.00%]
Highway Exit	24	0	24	100.00%	[86.20%, 100.00%]
Yield Bus Lane	24	0	24	100.00%	[86.20%, 100.00%]
Intersection	1,700	3	1,703	99.82%	[99.48%, 99.94%]
Crosswalk	975	0	975	100.00%	[99.61%, 100.00%]
Yield Road	844	1	845	99.88%	[99.33%, 99.98%]
Bus Stop	663	0	663	100.00%	[99.42%, 100.00%]
Traffic Light	622	2	624	99.68%	[98.84%, 99.91%]
Stop Sign	72	6	78	92.31%	[84.22%, 96.43%]
Roundabout	39	0	39	100.00%	[91.03%, 100.00%]
Avoid Bus Lane	28	11	39	71.79%	[56.22%, 83.46%]
Bridge	13	0	13	100.00%	[77.19%, 100.00%]
Highway Entry	13	0	13	100.00%	[77.19%, 100.00%]
Highway Exit	13	0	13	100.00%	[77.19%, 100.00%]

Time of Day

Daytime (All Cities, All Components)

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	9,408	112	9,520	98.82%	[98.59%, 99.02%]
Traffic Light	5,269	228	5,497	95.85%	[95.29%, 96.35%]
Crosswalk	3,700	2	3,702	99.95%	[99.80%, 99.99%]
Yield Bikelane	2,691	3	2,694	99.89%	[99.67%, 99.96%]
Yield Road	1,856	4	1,860	99.78%	[99.45%, 99.92%]
Digital Speed Sign	1,344	0	1,344	100.00%	[99.71%, 100.00%]
Roundabout	838	2	840	99.76%	[99.14%, 99.93%]
Yield Sidewalk	760	0	760	100.00%	[99.50%, 100.00%]
Bus Stop	755	1	756	99.87%	[99.25%, 99.98%]
Road Narrowing	749	7	756	99.07%	[98.10%, 99.55%]
Yield Tram	588	0	588	100.00%	[99.35%, 100.00%]
Yield Bus Lane	422	0	422	100.00%	[99.10%, 100.00%]
Bridge	415	5	420	98.81%	[97.24%, 99.49%]
Avoid Bus Lane	252	0	252	100.00%	[98.50%, 100.00%]
Priority Sign	248	4	252	98.41%	[95.99%, 99.38%]
Dynamic Barrier	169	1	170	99.41%	[96.74%, 99.90%]
School Zone	168	0	168	100.00%	[97.76%, 100.00%]
Living Street	86	0	86	100.00%	[95.72%, 100.00%]
Highway Entry	84	0	84	100.00%	[95.63%, 100.00%]
Highway Exit	84	0	84	100.00%	[95.63%, 100.00%]
U Turn	83	1	84	98.81%	[93.56%, 99.79%]
Intersection	10,174	158	10,332	98.47%	[98.22%, 98.69%]
Crosswalk	8,336	61	8,397	99.27%	[99.07%, 99.43%]
Traffic Light	6,711	209	6,920	96.98%	[96.55%, 97.36%]
Bus Stop	2,717	1	2,718	99.96%	[99.79%, 99.99%]
Yield Road	1,444	30	1,474	97.96%	[97.11%, 98.57%]
Roundabout	652	29	681	95.74%	[93.95%, 97.02%]
Stop Sign	415	37	452	91.81%	[88.92%, 94.00%]
Digital Speed Sign	322	17	339	94.99%	[92.12%, 96.85%]
Avoid Bus Lane	114	0	114	100.00%	[96.74%, 100.00%]
Yield Bikelane	113	0	113	100.00%	[96.71%, 100.00%]
Highway Entry	111	2	113	98.23%	[93.78%, 99.51%]
Highway Exit	87	26	113	76.99%	[68.42%, 83.79%]
Intersection	1,606	6	1,612	99.63%	[99.19%, 99.83%]
Bus Stop	1,240	0	1,240	100.00%	[99.69%, 100.00%]
Traffic Light	1,086	30	1,116	97.31%	[96.19%, 98.11%]
Crosswalk	1,020	3	1,023	99.71%	[99.14%, 99.90%]
Yield Bikelane	929	1	930	99.89%	[99.39%, 99.98%]
Yield Road	620	0	620	100.00%	[99.38%, 100.00%]
Digital Speed Sign	215	33	248	86.69%	[81.90%, 90.37%]
Roundabout	184	2	186	98.92%	[96.16%, 99.70%]
Yield Sidewalk	124	0	124	100.00%	[97.00%, 100.00%]
Road Narrowing	90	1	91	98.90%	[94.04%, 99.81%]
Highway Exit	62	0	62	100.00%	[94.17%, 100.00%]
Highway Entry	31	0	31	100.00%	[88.97%, 100.00%]
Intersection	9,582	37	9,619	99.62%	[99.47%, 99.72%]
Traffic Light	2,840	40	2,880	98.61%	[98.11%, 98.98%]
Priority	2,366	2	2,368	99.92%	[99.69%, 99.98%]
Yield Sidewalk	1,950	1	1,951	99.95%	[99.71%, 99.99%]
Bus Stop	1,257	1	1,258	99.92%	[99.55%, 99.99%]
Digital sign	787	27	814	96.68%	[95.22%, 97.71%]
Crosswalk	591	1	592	99.83%	[99.05%, 99.97%]
Roundabout	444	0	444	100.00%	[99.14%, 100.00%]
Road Narrowing	219	3	222	98.65%	[96.10%, 99.54%]
Stop Sign	125	3	128	97.66%	[93.34%, 99.20%]
Highway Entry	74	0	74	100.00%	[95.07%, 100.00%]
Highway Exit	74	0	74	100.00%	[95.07%, 100.00%]
Living Street	27	1	28	96.43%	[82.29%, 99.37%]
Crosswalk	12,228	27	12,255	99.78%	[99.68%, 99.85%]
Intersection	5,016	19	5,035	99.62%	[99.41%, 99.76%]
Yield Road	3,976	14	3,990	99.65%	[99.41%, 99.79%]
Traffic Light	3,647	58	3,705	98.43%	[97.98%, 98.79%]
Roundabout	2,743	12	2,755	99.56%	[99.24%, 99.75%]
Bus Stop	2,469	1	2,470	99.96%	[99.77%, 99.99%]
Stop Sign	843	12	855	98.60%	[97.56%, 99.20%]
Yield Bikelane	570	0	570	100.00%	[99.33%, 100.00%]
Road Narrowing	377	3	380	99.21%	[97.70%, 99.73%]
Bridge	285	0	285	100.00%	[98.67%, 100.00%]
Priority Sign	285	0	285	100.00%	[98.67%, 100.00%]
School Zone	282	3	285	98.95%	[96.95%, 99.64%]
Yield Sidewalk	282	3	285	98.95%	[96.95%, 99.64%]
Highway Entry	190	0	190	100.00%	[98.02%, 100.00%]
Highway Exit	190	0	190	100.00%	[98.02%, 100.00%]
Avoid Bus Lane	187	3	190	98.42%	[95.46%, 99.46%]
Yield Bus Lane	187	3	190	98.42%	[95.46%, 99.46%]
Intersection	12,692	17	12,709	99.87%	[99.79%, 99.92%]
Crosswalk	7,264	13	7,277	99.82%	[99.69%, 99.90%]
Yield Road	6,295	10	6,305	99.84%	[99.71%, 99.91%]
Bus Stop	4,947	0	4,947	100.00%	[99.92%, 100.00%]
Traffic Light	4,630	26	4,656	99.44%	[99.18%, 99.62%]
Stop Sign	513	69	582	88.14%	[85.26%, 90.52%]
Roundabout	291	0	291	100.00%	[98.70%, 100.00%]
Avoid Bus Lane	233	58	291	80.07%	[75.10%, 84.25%]
Bridge	97	0	97	100.00%	[96.19%, 100.00%]
Highway Entry	97	0	97	100.00%	[96.19%, 100.00%]
Highway Exit	97	0	97	100.00%	[96.19%, 100.00%]

Aggregated Pass Rates (All Cities, All Components)

Download Detailed Data

Download by City CSV Download by Weather CSV Download by Time CSV

Study 8: Max Speed Analysis

Tesla has conducted an analysis of 708 randomly selected driving samples from the European engineering fleet. Each sample satisfies the following criteria:

The vehicle is driving in Europe
At least one other vehicle is detected ahead of or behind the ego vehicle, travelling in the same direction
The current maximum speed has been increased above the system-determined speed limit by vision max speed
The vehicle is on a controlled access road or city street
The vehicle speed exceeds the system-determined speed limit

For each sample, the ego vehicle's speed is compared against surrounding traffic in two ways:

Median comparison: The difference between the ego vehicle's speed and the median speed of surrounding same-direction traffic
Fastest comparison: The difference between the ego vehicle's speed and the speed of the fastest surrounding vehicle

Low-speed scenarios (below 25 km/h for either the ego vehicle or the median surrounding traffic) are excluded from the analysis, resulting in 680 samples for the chart below.

Exemption Relevance: This study directly supports Exemption Request 5 (contextual speed determination) by demonstrating that when vision max speed increases the current maximum speed above the system-determined speed limit, FSD (Supervised) maintains a speed consistent with the surrounding traffic flow rather than driving faster than other vehicles on the road.

Study Takeaways

In 98% of high-speed cases, FSD (Supervised) travels at or below the median speed of surrounding traffic, even when the current maximum speed is set above the system-determined speed limit by vision max speed.
In 12 out of 680 high-speed samples, the ego vehicle slightly exceeds the median traffic speed. In none of those cases does the ego vehicle exceed the speed of the fastest surrounding vehicle.
This data demonstrates that vision max speed corrects for incorrect speed limit determination without introducing excessive speeding behaviour. The vehicle speed remains consistent with the surrounding traffic flow, reducing the risk of driving significantly slower than surrounding vehicles due to an incorrect system-determined speed limit.

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	2,468	18	2,486	99.28%	[98.86%, 99.54%]
Traffic Light	1,333	54	1,387	96.11%	[94.95%, 97.00%]
Crosswalk	968	0	968	100.00%	[99.60%, 100.00%]
Yield Bikelane	704	0	704	100.00%	[99.46%, 100.00%]
Yield Road	484	0	484	100.00%	[99.21%, 100.00%]
Digital Speed Sign	352	0	352	100.00%	[98.92%, 100.00%]
Roundabout	220	0	220	100.00%	[98.28%, 100.00%]
Bus Stop	198	0	198	100.00%	[98.10%, 100.00%]
Road Narrowing	198	0	198	100.00%	[98.10%, 100.00%]
Yield Sidewalk	198	0	198	100.00%	[98.10%, 100.00%]
Yield Tram	154	0	154	100.00%	[97.57%, 100.00%]
Yield Bus Lane	110	0	110	100.00%	[96.63%, 100.00%]
Bridge	109	1	110	99.09%	[95.03%, 99.84%]
Avoid Bus Lane	66	0	66	100.00%	[94.50%, 100.00%]
Priority Sign	66	0	66	100.00%	[94.50%, 100.00%]
Dynamic Barrier	44	0	44	100.00%	[91.97%, 100.00%]
School Zone	44	0	44	100.00%	[91.97%, 100.00%]
Highway Entry	22	0	22	100.00%	[85.13%, 100.00%]
Highway Exit	22	0	22	100.00%	[85.13%, 100.00%]
Living Street	22	0	22	100.00%	[85.13%, 100.00%]
U Turn	22	0	22	100.00%	[85.13%, 100.00%]
Intersection	2,063	30	2,093	98.57%	[97.96%, 98.99%]
Crosswalk	1,690	12	1,702	99.29%	[98.77%, 99.60%]
Traffic Light	1,359	44	1,403	96.86%	[95.82%, 97.66%]
Bus Stop	552	0	552	100.00%	[99.31%, 100.00%]
Yield Road	289	10	299	96.66%	[93.95%, 98.17%]
Roundabout	134	4	138	97.10%	[92.78%, 98.87%]
Stop Sign	86	6	92	93.48%	[86.49%, 96.98%]
Digital Speed Sign	66	3	69	95.65%	[87.98%, 98.51%]
Avoid Bus Lane	23	0	23	100.00%	[85.69%, 100.00%]
Yield Bikelane	23	0	23	100.00%	[85.69%, 100.00%]
Highway Entry	21	2	23	91.30%	[73.20%, 97.58%]
Highway Exit	17	6	23	73.91%	[53.53%, 87.45%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Digital Speed Sign	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Road Narrowing	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Bikelane	0	0	0	nan%	[0.00%, 0.00%]
Yield Road	0	0	0	nan%	[0.00%, 0.00%]
Yield Sidewalk	0	0	0	nan%	[0.00%, 0.00%]
Intersection	3,245	4	3,249	99.88%	[99.68%, 99.95%]
Traffic Light	1,018	17	1,035	98.36%	[97.39%, 98.97%]
Priority	800	0	800	100.00%	[99.52%, 100.00%]
Yield Sidewalk	657	1	658	99.85%	[99.14%, 99.97%]
Bus Stop	425	0	425	100.00%	[99.10%, 100.00%]
Digital sign	260	15	275	94.55%	[91.20%, 96.67%]
Crosswalk	200	0	200	100.00%	[98.12%, 100.00%]
Roundabout	150	0	150	100.00%	[97.50%, 100.00%]
Road Narrowing	74	1	75	98.67%	[92.83%, 99.76%]
Stop Sign	44	2	46	95.65%	[85.47%, 98.80%]
Highway Entry	25	0	25	100.00%	[86.68%, 100.00%]
Highway Exit	25	0	25	100.00%	[86.68%, 100.00%]
Living Street	9	0	9	100.00%	[70.09%, 100.00%]
Crosswalk	4,249	8	4,257	99.81%	[99.63%, 99.90%]
Intersection	1,742	7	1,749	99.60%	[99.18%, 99.81%]
Yield Road	1,382	4	1,386	99.71%	[99.26%, 99.89%]
Traffic Light	1,264	23	1,287	98.21%	[97.33%, 98.81%]
Roundabout	955	2	957	99.79%	[99.24%, 99.94%]
Bus Stop	858	0	858	100.00%	[99.55%, 100.00%]
Stop Sign	293	4	297	98.65%	[96.59%, 99.48%]
Yield Bikelane	198	0	198	100.00%	[98.10%, 100.00%]
Road Narrowing	131	1	132	99.24%	[95.83%, 99.87%]
Bridge	99	0	99	100.00%	[96.26%, 100.00%]
Priority Sign	99	0	99	100.00%	[96.26%, 100.00%]
School Zone	97	2	99	97.98%	[92.93%, 99.44%]
Yield Sidewalk	97	2	99	97.98%	[92.93%, 99.44%]
Highway Entry	66	0	66	100.00%	[94.50%, 100.00%]
Highway Exit	65	1	66	98.48%	[91.90%, 99.73%]
Avoid Bus Lane	63	3	66	95.45%	[87.47%, 98.44%]
Yield Bus Lane	63	3	66	95.45%	[87.47%, 98.44%]
Intersection	1,179	0	1,179	100.00%	[99.68%, 100.00%]
Crosswalk	675	0	675	100.00%	[99.43%, 100.00%]
Yield Road	585	0	585	100.00%	[99.35%, 100.00%]
Bus Stop	459	0	459	100.00%	[99.17%, 100.00%]
Traffic Light	430	2	432	99.54%	[98.33%, 99.87%]
Stop Sign	49	5	54	90.74%	[80.09%, 95.98%]
Roundabout	27	0	27	100.00%	[87.54%, 100.00%]
Avoid Bus Lane	24	3	27	88.89%	[71.94%, 96.15%]
Bridge	9	0	9	100.00%	[70.09%, 100.00%]
Highway Entry	9	0	9	100.00%	[70.09%, 100.00%]
Highway Exit	9	0	9	100.00%	[70.09%, 100.00%]

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	4,248	70	4,318	98.38%	[97.96%, 98.71%]
Traffic Light	2,392	130	2,522	94.85%	[93.91%, 95.64%]
Crosswalk	1,673	3	1,676	99.82%	[99.48%, 99.94%]
Yield Bikelane	1,220	2	1,222	99.84%	[99.41%, 99.96%]
Yield Road	842	2	844	99.76%	[99.14%, 99.93%]
Digital Speed Sign	608	0	608	100.00%	[99.37%, 100.00%]
Roundabout	378	2	380	99.47%	[98.10%, 99.86%]
Yield Sidewalk	344	0	344	100.00%	[98.90%, 100.00%]
Bus Stop	342	0	342	100.00%	[98.89%, 100.00%]
Road Narrowing	338	4	342	98.83%	[97.03%, 99.54%]
Yield Tram	266	0	266	100.00%	[98.58%, 100.00%]
Yield Bus Lane	190	0	190	100.00%	[98.02%, 100.00%]
Bridge	189	1	190	99.47%	[97.08%, 99.91%]
Avoid Bus Lane	114	0	114	100.00%	[96.74%, 100.00%]
Priority Sign	113	1	114	99.12%	[95.20%, 99.84%]
Dynamic Barrier	76	0	76	100.00%	[95.19%, 100.00%]
School Zone	76	0	76	100.00%	[95.19%, 100.00%]
Living Street	40	0	40	100.00%	[91.24%, 100.00%]
Highway Entry	38	0	38	100.00%	[90.82%, 100.00%]
Highway Exit	38	0	38	100.00%	[90.82%, 100.00%]
U Turn	38	0	38	100.00%	[90.82%, 100.00%]
Intersection	271	2	273	99.27%	[97.37%, 99.80%]
Crosswalk	220	2	222	99.10%	[96.78%, 99.75%]
Traffic Light	178	5	183	97.27%	[93.76%, 98.83%]
Bus Stop	72	0	72	100.00%	[94.93%, 100.00%]
Yield Road	39	0	39	100.00%	[91.03%, 100.00%]
Roundabout	17	1	18	94.44%	[74.24%, 99.01%]
Stop Sign	10	2	12	83.33%	[55.20%, 95.30%]
Digital Speed Sign	7	2	9	77.78%	[45.26%, 93.68%]
Avoid Bus Lane	3	0	3	100.00%	[43.85%, 100.00%]
Highway Entry	3	0	3	100.00%	[43.85%, 100.00%]
Highway Exit	3	0	3	100.00%	[43.85%, 100.00%]
Yield Bikelane	3	0	3	100.00%	[43.85%, 100.00%]
Intersection	779	1	780	99.87%	[99.28%, 99.98%]
Bus Stop	600	0	600	100.00%	[99.36%, 100.00%]
Traffic Light	528	12	540	97.78%	[96.16%, 98.72%]
Crosswalk	492	3	495	99.39%	[98.23%, 99.79%]
Yield Bikelane	449	1	450	99.78%	[98.75%, 99.96%]
Yield Road	300	0	300	100.00%	[98.74%, 100.00%]
Digital Speed Sign	106	14	120	88.33%	[81.37%, 92.92%]
Roundabout	89	1	90	98.89%	[93.97%, 99.80%]
Yield Sidewalk	60	0	60	100.00%	[93.98%, 100.00%]
Road Narrowing	44	0	44	100.00%	[91.97%, 100.00%]
Highway Exit	30	0	30	100.00%	[88.65%, 100.00%]
Highway Entry	15	0	15	100.00%	[79.61%, 100.00%]
Intersection	3,503	7	3,510	99.80%	[99.59%, 99.90%]
Traffic Light	1,068	12	1,080	98.89%	[98.07%, 99.36%]
Priority	864	0	864	100.00%	[99.56%, 100.00%]
Yield Sidewalk	709	1	710	99.86%	[99.21%, 99.98%]
Bus Stop	459	0	459	100.00%	[99.17%, 100.00%]
Digital sign	283	14	297	95.29%	[92.24%, 97.17%]
Crosswalk	216	0	216	100.00%	[98.25%, 100.00%]
Roundabout	162	0	162	100.00%	[97.68%, 100.00%]
Road Narrowing	80	1	81	98.77%	[93.33%, 99.78%]
Stop Sign	47	1	48	97.92%	[89.10%, 99.63%]
Highway Entry	27	0	27	100.00%	[87.54%, 100.00%]
Highway Exit	27	0	27	100.00%	[87.54%, 100.00%]
Living Street	8	0	8	100.00%	[67.56%, 100.00%]
Crosswalk	3,478	5	3,483	99.86%	[99.66%, 99.94%]
Intersection	1,424	7	1,431	99.51%	[98.99%, 99.76%]
Yield Road	1,129	5	1,134	99.56%	[98.97%, 99.81%]
Traffic Light	1,036	17	1,053	98.39%	[97.43%, 98.99%]
Roundabout	778	5	783	99.36%	[98.51%, 99.73%]
Bus Stop	701	1	702	99.86%	[99.20%, 99.97%]
Stop Sign	239	4	243	98.35%	[95.84%, 99.36%]
Yield Bikelane	162	0	162	100.00%	[97.68%, 100.00%]
Road Narrowing	108	0	108	100.00%	[96.57%, 100.00%]
Bridge	81	0	81	100.00%	[95.47%, 100.00%]
Priority Sign	81	0	81	100.00%	[95.47%, 100.00%]
School Zone	80	1	81	98.77%	[93.33%, 99.78%]
Yield Sidewalk	80	1	81	98.77%	[93.33%, 99.78%]
Highway Entry	54	0	54	100.00%	[93.36%, 100.00%]
Highway Exit	54	0	54	100.00%	[93.36%, 100.00%]
Avoid Bus Lane	52	2	54	96.30%	[87.46%, 98.98%]
Yield Bus Lane	52	2	54	96.30%	[87.46%, 98.98%]
Intersection	784	2	786	99.75%	[99.08%, 99.93%]
Crosswalk	450	0	450	100.00%	[99.15%, 100.00%]
Yield Road	390	0	390	100.00%	[99.02%, 100.00%]
Bus Stop	306	0	306	100.00%	[98.76%, 100.00%]
Traffic Light	286	2	288	99.31%	[97.50%, 99.81%]
Stop Sign	33	3	36	91.67%	[78.17%, 97.13%]
Roundabout	18	0	18	100.00%	[82.41%, 100.00%]
Avoid Bus Lane	15	3	18	83.33%	[60.78%, 94.16%]
Bridge	6	0	6	100.00%	[60.97%, 100.00%]
Highway Entry	6	0	6	100.00%	[60.97%, 100.00%]
Highway Exit	6	0	6	100.00%	[60.97%, 100.00%]

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	2,240	24	2,264	98.94%	[98.43%, 99.29%]
Traffic Light	1,270	54	1,324	95.92%	[94.72%, 96.86%]
Crosswalk	881	1	882	99.89%	[99.36%, 99.98%]
Yield Bikelane	640	0	640	100.00%	[99.40%, 100.00%]
Yield Road	442	2	444	99.55%	[98.37%, 99.88%]
Digital Speed Sign	320	0	320	100.00%	[98.81%, 100.00%]
Roundabout	200	0	200	100.00%	[98.12%, 100.00%]
Yield Sidewalk	182	0	182	100.00%	[97.93%, 100.00%]
Bus Stop	179	1	180	99.44%	[96.92%, 99.90%]
Road Narrowing	179	1	180	99.44%	[96.92%, 99.90%]
Yield Tram	140	0	140	100.00%	[97.33%, 100.00%]
Yield Bus Lane	102	0	102	100.00%	[96.37%, 100.00%]
Bridge	98	2	100	98.00%	[93.00%, 99.45%]
Avoid Bus Lane	60	0	60	100.00%	[93.98%, 100.00%]
Priority Sign	59	1	60	98.33%	[91.14%, 99.71%]
Dynamic Barrier	42	0	42	100.00%	[91.62%, 100.00%]
School Zone	40	0	40	100.00%	[91.24%, 100.00%]
Highway Entry	20	0	20	100.00%	[83.89%, 100.00%]
Highway Exit	20	0	20	100.00%	[83.89%, 100.00%]
Living Street	19	1	20	95.00%	[76.39%, 99.11%]
U Turn	18	2	20	90.00%	[69.90%, 97.21%]
Intersection	10,092	149	10,241	98.55%	[98.29%, 98.76%]
Crosswalk	8,270	53	8,323	99.36%	[99.17%, 99.51%]
Traffic Light	6,651	208	6,859	96.97%	[96.53%, 97.35%]
Bus Stop	2,693	1	2,694	99.96%	[99.79%, 99.99%]
Yield Road	1,434	27	1,461	98.15%	[97.32%, 98.73%]
Roundabout	638	37	675	94.52%	[92.54%, 96.00%]
Stop Sign	406	42	448	90.62%	[87.57%, 92.99%]
Digital Speed Sign	316	20	336	94.05%	[90.99%, 96.11%]
Avoid Bus Lane	113	0	113	100.00%	[96.71%, 100.00%]
Highway Entry	112	0	112	100.00%	[96.68%, 100.00%]
Yield Bikelane	112	0	112	100.00%	[96.68%, 100.00%]
Highway Exit	83	29	112	74.11%	[65.29%, 81.32%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Digital Speed Sign	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Road Narrowing	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Bikelane	0	0	0	nan%	[0.00%, 0.00%]
Yield Road	0	0	0	nan%	[0.00%, 0.00%]
Yield Sidewalk	0	0	0	nan%	[0.00%, 0.00%]
Intersection	2,711	19	2,730	99.30%	[98.92%, 99.55%]
Traffic Light	760	5	765	99.35%	[98.48%, 99.72%]
Priority	669	3	672	99.55%	[98.70%, 99.85%]
Yield Sidewalk	555	0	555	100.00%	[99.31%, 100.00%]
Bus Stop	356	1	357	99.72%	[98.43%, 99.95%]
Digital sign	225	6	231	97.40%	[94.45%, 98.80%]
Crosswalk	167	1	168	99.40%	[96.71%, 99.89%]
Roundabout	126	0	126	100.00%	[97.04%, 100.00%]
Road Narrowing	62	1	63	98.41%	[91.54%, 99.72%]
Stop Sign	34	0	34	100.00%	[89.85%, 100.00%]
Highway Entry	21	0	21	100.00%	[84.54%, 100.00%]
Highway Exit	21	0	21	100.00%	[84.54%, 100.00%]
Living Street	8	1	9	88.89%	[56.50%, 98.01%]
Crosswalk	3,857	13	3,870	99.66%	[99.43%, 99.80%]
Intersection	1,583	7	1,590	99.56%	[99.09%, 99.79%]
Yield Road	1,256	4	1,260	99.68%	[99.19%, 99.88%]
Traffic Light	1,148	22	1,170	98.12%	[97.17%, 98.76%]
Roundabout	867	3	870	99.66%	[98.99%, 99.88%]
Bus Stop	780	0	780	100.00%	[99.51%, 100.00%]
Stop Sign	264	6	270	97.78%	[95.24%, 98.98%]
Yield Bikelane	180	0	180	100.00%	[97.91%, 100.00%]
Road Narrowing	118	2	120	98.33%	[94.13%, 99.54%]
Bridge	90	0	90	100.00%	[95.91%, 100.00%]
Priority Sign	90	0	90	100.00%	[95.91%, 100.00%]
School Zone	90	0	90	100.00%	[95.91%, 100.00%]
Yield Sidewalk	90	0	90	100.00%	[95.91%, 100.00%]
Avoid Bus Lane	60	0	60	100.00%	[93.98%, 100.00%]
Highway Entry	60	0	60	100.00%	[93.98%, 100.00%]
Highway Exit	60	0	60	100.00%	[93.98%, 100.00%]
Yield Bus Lane	60	0	60	100.00%	[93.98%, 100.00%]
Intersection	13,874	14	13,888	99.90%	[99.83%, 99.94%]
Crosswalk	7,936	16	7,952	99.80%	[99.67%, 99.88%]
Yield Road	6,879	11	6,890	99.84%	[99.71%, 99.91%]
Bus Stop	5,406	0	5,406	100.00%	[99.93%, 100.00%]
Traffic Light	5,052	36	5,088	99.29%	[99.02%, 99.49%]
Stop Sign	548	88	636	86.16%	[83.26%, 88.63%]
Roundabout	318	0	318	100.00%	[98.81%, 100.00%]
Avoid Bus Lane	250	68	318	78.62%	[73.78%, 82.77%]
Bridge	106	0	106	100.00%	[96.50%, 100.00%]
Highway Entry	106	0	106	100.00%	[96.50%, 100.00%]
Highway Exit	106	0	106	100.00%	[96.50%, 100.00%]

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	7,802	94	7,896	98.81%	[98.55%, 99.03%]
Crosswalk	3,040	16	3,056	99.48%	[99.15%, 99.68%]
Yield Bikelane	2,177	6	2,183	99.73%	[99.40%, 99.87%]
Yield Road	1,510	4	1,514	99.74%	[99.32%, 99.90%]
Digital Speed Sign	1,135	1	1,136	99.91%	[99.50%, 99.98%]
Roundabout	663	3	666	99.55%	[98.68%, 99.85%]
Yield Sidewalk	636	0	636	100.00%	[99.40%, 100.00%]
Bus Stop	632	2	634	99.68%	[98.86%, 99.91%]
Road Narrowing	625	6	631	99.05%	[97.94%, 99.56%]
Yield Tram	495	0	495	100.00%	[99.23%, 100.00%]
Traffic Light	429	32	461	93.06%	[90.37%, 95.04%]
Yield Bus Lane	351	1	352	99.72%	[98.41%, 99.95%]
Bridge	346	7	353	98.02%	[95.96%, 99.04%]
Avoid Bus Lane	210	1	211	99.53%	[97.36%, 99.92%]
Priority Sign	209	2	211	99.05%	[96.61%, 99.74%]
School Zone	141	0	141	100.00%	[97.35%, 100.00%]
Dynamic Barrier	139	2	141	98.58%	[94.98%, 99.61%]
Highway Entry	71	0	71	100.00%	[94.87%, 100.00%]
U Turn	71	0	71	100.00%	[94.87%, 100.00%]
Highway Exit	70	1	71	98.59%	[92.44%, 99.75%]
Living Street	70	1	71	98.59%	[92.44%, 99.75%]
Intersection	3,850	63	3,913	98.39%	[97.95%, 98.74%]
Crosswalk	3,164	18	3,182	99.43%	[99.11%, 99.64%]
Traffic Light	2,543	80	2,623	96.95%	[96.22%, 97.54%]
Bus Stop	1,031	1	1,032	99.90%	[99.45%, 99.98%]
Yield Road	553	6	559	98.93%	[97.68%, 99.51%]
Roundabout	248	10	258	96.12%	[93.01%, 97.88%]
Stop Sign	162	10	172	94.19%	[89.63%, 96.81%]
Digital Speed Sign	123	6	129	95.35%	[90.22%, 97.85%]
Avoid Bus Lane	43	0	43	100.00%	[91.80%, 100.00%]
Highway Entry	43	0	43	100.00%	[91.80%, 100.00%]
Yield Bikelane	43	0	43	100.00%	[91.80%, 100.00%]
Highway Exit	31	12	43	72.09%	[57.31%, 83.25%]
Intersection	775	5	780	99.36%	[98.51%, 99.73%]
Bus Stop	600	0	600	100.00%	[99.36%, 100.00%]
Traffic Light	524	16	540	97.04%	[95.24%, 98.17%]
Crosswalk	495	0	495	100.00%	[99.23%, 100.00%]
Yield Bikelane	450	0	450	100.00%	[99.15%, 100.00%]
Yield Road	300	0	300	100.00%	[98.74%, 100.00%]
Digital Speed Sign	101	19	120	84.17%	[76.59%, 89.62%]
Roundabout	89	1	90	98.89%	[93.97%, 99.80%]
Yield Sidewalk	60	0	60	100.00%	[93.98%, 100.00%]
Road Narrowing	43	1	44	97.73%	[88.19%, 99.60%]
Highway Exit	30	0	30	100.00%	[88.65%, 100.00%]
Highway Entry	15	0	15	100.00%	[79.61%, 100.00%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Digital sign	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Living Street	0	0	0	nan%	[0.00%, 0.00%]
Priority	0	0	0	nan%	[0.00%, 0.00%]
Road Narrowing	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Stop Sign	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Sidewalk	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	387	0	387	100.00%	[99.02%, 100.00%]
Intersection	159	0	159	100.00%	[97.64%, 100.00%]
Yield Road	126	0	126	100.00%	[97.04%, 100.00%]
Traffic Light	114	3	117	97.44%	[92.73%, 99.12%]
Roundabout	87	0	87	100.00%	[95.77%, 100.00%]
Bus Stop	77	1	78	98.72%	[93.09%, 99.77%]
Stop Sign	27	0	27	100.00%	[87.54%, 100.00%]
Yield Bikelane	18	0	18	100.00%	[82.41%, 100.00%]
Road Narrowing	12	0	12	100.00%	[75.75%, 100.00%]
Bridge	9	0	9	100.00%	[70.09%, 100.00%]
Priority Sign	9	0	9	100.00%	[70.09%, 100.00%]
School Zone	9	0	9	100.00%	[70.09%, 100.00%]
Yield Sidewalk	9	0	9	100.00%	[70.09%, 100.00%]
Highway Entry	6	0	6	100.00%	[60.97%, 100.00%]
Highway Exit	6	0	6	100.00%	[60.97%, 100.00%]
Avoid Bus Lane	5	1	6	83.33%	[43.65%, 96.99%]
Yield Bus Lane	5	1	6	83.33%	[43.65%, 96.99%]
Intersection	784	2	786	99.75%	[99.08%, 99.93%]
Crosswalk	449	1	450	99.78%	[98.75%, 99.96%]
Yield Road	389	1	390	99.74%	[98.56%, 99.95%]
Bus Stop	306	0	306	100.00%	[98.76%, 100.00%]
Traffic Light	287	1	288	99.65%	[98.06%, 99.94%]
Stop Sign	28	8	36	77.78%	[61.92%, 88.28%]
Roundabout	18	0	18	100.00%	[82.41%, 100.00%]
Avoid Bus Lane	15	3	18	83.33%	[60.78%, 94.16%]
Bridge	6	0	6	100.00%	[60.97%, 100.00%]
Highway Entry	6	0	6	100.00%	[60.97%, 100.00%]
Highway Exit	6	0	6	100.00%	[60.97%, 100.00%]

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	331	8	339	97.64%	[95.41%, 98.80%]
Traffic Light	193	5	198	97.47%	[94.23%, 98.92%]
Crosswalk	132	0	132	100.00%	[97.17%, 100.00%]
Yield Bikelane	95	1	96	98.96%	[94.33%, 99.82%]
Yield Road	65	1	66	98.48%	[91.90%, 99.73%]
Digital Speed Sign	48	0	48	100.00%	[92.59%, 100.00%]
Roundabout	30	0	30	100.00%	[88.65%, 100.00%]
Bus Stop	27	0	27	100.00%	[87.54%, 100.00%]
Road Narrowing	27	0	27	100.00%	[87.54%, 100.00%]
Yield Sidewalk	27	0	27	100.00%	[87.54%, 100.00%]
Yield Tram	21	0	21	100.00%	[84.54%, 100.00%]
Bridge	15	0	15	100.00%	[79.61%, 100.00%]
Yield Bus Lane	15	0	15	100.00%	[79.61%, 100.00%]
Avoid Bus Lane	9	0	9	100.00%	[70.09%, 100.00%]
Priority Sign	9	0	9	100.00%	[70.09%, 100.00%]
Dynamic Barrier	6	0	6	100.00%	[60.97%, 100.00%]
School Zone	6	0	6	100.00%	[60.97%, 100.00%]
Highway Entry	3	0	3	100.00%	[43.85%, 100.00%]
Highway Exit	3	0	3	100.00%	[43.85%, 100.00%]
Living Street	3	0	3	100.00%	[43.85%, 100.00%]
U Turn	3	0	3	100.00%	[43.85%, 100.00%]
Avoid Bus Lane	0	0	0	nan%	[0.00%, 0.00%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Digital Speed Sign	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Stop Sign	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Bikelane	0	0	0	nan%	[0.00%, 0.00%]
Yield Road	0	0	0	nan%	[0.00%, 0.00%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Digital Speed Sign	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Road Narrowing	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Bikelane	0	0	0	nan%	[0.00%, 0.00%]
Yield Road	0	0	0	nan%	[0.00%, 0.00%]
Yield Sidewalk	0	0	0	nan%	[0.00%, 0.00%]
Intersection	387	3	390	99.23%	[97.76%, 99.74%]
Traffic Light	132	3	135	97.78%	[93.67%, 99.24%]
Priority	95	1	96	98.96%	[94.33%, 99.82%]
Yield Sidewalk	78	0	78	100.00%	[95.31%, 100.00%]
Bus Stop	51	0	51	100.00%	[93.00%, 100.00%]
Digital sign	31	2	33	93.94%	[80.39%, 98.32%]
Crosswalk	24	0	24	100.00%	[86.20%, 100.00%]
Roundabout	18	0	18	100.00%	[82.41%, 100.00%]
Road Narrowing	9	0	9	100.00%	[70.09%, 100.00%]
Stop Sign	6	0	6	100.00%	[60.97%, 100.00%]
Highway Entry	3	0	3	100.00%	[43.85%, 100.00%]
Highway Exit	3	0	3	100.00%	[43.85%, 100.00%]
Living Street	0	0	0	nan%	[0.00%, 0.00%]
Avoid Bus Lane	0	0	0	nan%	[0.00%, 0.00%]
Bridge	0	0	0	nan%	[0.00%, 0.00%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Priority Sign	0	0	0	nan%	[0.00%, 0.00%]
Road Narrowing	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
School Zone	0	0	0	nan%	[0.00%, 0.00%]
Stop Sign	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Bikelane	0	0	0	nan%	[0.00%, 0.00%]
Yield Bus Lane	0	0	0	nan%	[0.00%, 0.00%]
Yield Road	0	0	0	nan%	[0.00%, 0.00%]
Yield Sidewalk	0	0	0	nan%	[0.00%, 0.00%]
Avoid Bus Lane	0	0	0	nan%	[0.00%, 0.00%]
Bridge	0	0	0	nan%	[0.00%, 0.00%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Stop Sign	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Road	0	0	0	nan%	[0.00%, 0.00%]

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	893	11	904	98.78%	[97.83%, 99.32%]
Traffic Light	508	20	528	96.21%	[94.22%, 97.53%]
Crosswalk	350	2	352	99.43%	[97.95%, 99.84%]
Yield Bikelane	256	0	256	100.00%	[98.52%, 100.00%]
Yield Road	176	0	176	100.00%	[97.86%, 100.00%]
Digital Speed Sign	128	0	128	100.00%	[97.09%, 100.00%]
Roundabout	80	0	80	100.00%	[95.42%, 100.00%]
Bus Stop	72	0	72	100.00%	[94.93%, 100.00%]
Road Narrowing	72	0	72	100.00%	[94.93%, 100.00%]
Yield Sidewalk	72	0	72	100.00%	[94.93%, 100.00%]
Yield Tram	56	0	56	100.00%	[93.58%, 100.00%]
Bridge	40	0	40	100.00%	[91.24%, 100.00%]
Yield Bus Lane	40	0	40	100.00%	[91.24%, 100.00%]
Avoid Bus Lane	24	0	24	100.00%	[86.20%, 100.00%]
Priority Sign	24	0	24	100.00%	[86.20%, 100.00%]
Dynamic Barrier	16	0	16	100.00%	[80.64%, 100.00%]
School Zone	16	0	16	100.00%	[80.64%, 100.00%]
Highway Entry	8	0	8	100.00%	[67.56%, 100.00%]
Highway Exit	8	0	8	100.00%	[67.56%, 100.00%]
Living Street	8	0	8	100.00%	[67.56%, 100.00%]
U Turn	8	0	8	100.00%	[67.56%, 100.00%]
Intersection	1,169	14	1,183	98.82%	[98.02%, 99.29%]
Crosswalk	960	2	962	99.79%	[99.25%, 99.94%]
Traffic Light	773	20	793	97.48%	[96.14%, 98.36%]
Bus Stop	312	0	312	100.00%	[98.78%, 100.00%]
Yield Road	166	3	169	98.22%	[94.91%, 99.39%]
Roundabout	74	4	78	94.87%	[87.54%, 97.99%]
Stop Sign	49	3	52	94.23%	[84.36%, 98.02%]
Digital Speed Sign	37	2	39	94.87%	[83.11%, 98.58%]
Avoid Bus Lane	13	0	13	100.00%	[77.19%, 100.00%]
Highway Entry	13	0	13	100.00%	[77.19%, 100.00%]
Yield Bikelane	13	0	13	100.00%	[77.19%, 100.00%]
Highway Exit	10	3	13	76.92%	[49.74%, 91.82%]
Intersection	52	0	52	100.00%	[93.12%, 100.00%]
Bus Stop	40	0	40	100.00%	[91.24%, 100.00%]
Traffic Light	36	0	36	100.00%	[90.36%, 100.00%]
Crosswalk	33	0	33	100.00%	[89.57%, 100.00%]
Yield Bikelane	30	0	30	100.00%	[88.65%, 100.00%]
Yield Road	20	0	20	100.00%	[83.89%, 100.00%]
Digital Speed Sign	8	0	8	100.00%	[67.56%, 100.00%]
Roundabout	6	0	6	100.00%	[60.97%, 100.00%]
Yield Sidewalk	4	0	4	100.00%	[51.01%, 100.00%]
Road Narrowing	3	0	3	100.00%	[43.85%, 100.00%]
Highway Exit	2	0	2	100.00%	[34.24%, 100.00%]
Highway Entry	1	0	1	100.00%	[20.65%, 100.00%]
Intersection	1,039	1	1,040	99.90%	[99.46%, 99.98%]
Traffic Light	358	2	360	99.44%	[98.00%, 99.85%]
Priority	255	1	256	99.61%	[97.82%, 99.93%]
Yield Sidewalk	207	1	208	99.52%	[97.33%, 99.92%]
Bus Stop	136	0	136	100.00%	[97.25%, 100.00%]
Digital sign	76	12	88	86.36%	[77.66%, 92.02%]
Crosswalk	64	0	64	100.00%	[94.34%, 100.00%]
Roundabout	48	0	48	100.00%	[92.59%, 100.00%]
Road Narrowing	24	0	24	100.00%	[86.20%, 100.00%]
Stop Sign	16	0	16	100.00%	[80.64%, 100.00%]
Highway Entry	8	0	8	100.00%	[67.56%, 100.00%]
Highway Exit	8	0	8	100.00%	[67.56%, 100.00%]
Living Street	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	258	0	258	100.00%	[98.53%, 100.00%]
Intersection	104	2	106	98.11%	[93.38%, 99.48%]
Yield Road	84	0	84	100.00%	[95.63%, 100.00%]
Traffic Light	76	2	78	97.44%	[91.12%, 99.29%]
Roundabout	58	0	58	100.00%	[93.79%, 100.00%]
Bus Stop	52	0	52	100.00%	[93.12%, 100.00%]
Stop Sign	16	2	18	88.89%	[67.20%, 96.90%]
Yield Bikelane	12	0	12	100.00%	[75.75%, 100.00%]
Road Narrowing	8	0	8	100.00%	[67.56%, 100.00%]
Bridge	6	0	6	100.00%	[60.97%, 100.00%]
Priority Sign	6	0	6	100.00%	[60.97%, 100.00%]
School Zone	6	0	6	100.00%	[60.97%, 100.00%]
Yield Sidewalk	6	0	6	100.00%	[60.97%, 100.00%]
Highway Entry	4	0	4	100.00%	[51.01%, 100.00%]
Highway Exit	4	0	4	100.00%	[51.01%, 100.00%]
Avoid Bus Lane	3	1	4	75.00%	[30.06%, 95.44%]
Yield Bus Lane	3	1	4	75.00%	[30.06%, 95.44%]
Intersection	1,831	3	1,834	99.84%	[99.52%, 99.94%]
Crosswalk	1,047	3	1,050	99.71%	[99.16%, 99.90%]
Yield Road	909	1	910	99.89%	[99.38%, 99.98%]
Bus Stop	714	0	714	100.00%	[99.46%, 100.00%]
Traffic Light	663	9	672	98.66%	[97.47%, 99.29%]
Stop Sign	74	10	84	88.10%	[79.46%, 93.40%]
Roundabout	42	0	42	100.00%	[91.62%, 100.00%]
Avoid Bus Lane	31	11	42	73.81%	[58.93%, 84.70%]
Bridge	14	0	14	100.00%	[78.47%, 100.00%]
Highway Entry	14	0	14	100.00%	[78.47%, 100.00%]
Highway Exit	14	0	14	100.00%	[78.47%, 100.00%]

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	891	13	904	98.56%	[97.56%, 99.16%]
Traffic Light	499	29	528	94.51%	[92.22%, 96.15%]
Crosswalk	349	3	352	99.15%	[97.52%, 99.71%]
Yield Bikelane	256	0	256	100.00%	[98.52%, 100.00%]
Yield Road	174	2	176	98.86%	[95.95%, 99.69%]
Digital Speed Sign	128	0	128	100.00%	[97.09%, 100.00%]
Roundabout	80	0	80	100.00%	[95.42%, 100.00%]
Bus Stop	72	0	72	100.00%	[94.93%, 100.00%]
Road Narrowing	72	0	72	100.00%	[94.93%, 100.00%]
Yield Sidewalk	72	0	72	100.00%	[94.93%, 100.00%]
Yield Tram	56	0	56	100.00%	[93.58%, 100.00%]
Bridge	40	0	40	100.00%	[91.24%, 100.00%]
Yield Bus Lane	40	0	40	100.00%	[91.24%, 100.00%]
Avoid Bus Lane	24	0	24	100.00%	[86.20%, 100.00%]
Priority Sign	24	0	24	100.00%	[86.20%, 100.00%]
Dynamic Barrier	16	0	16	100.00%	[80.64%, 100.00%]
School Zone	16	0	16	100.00%	[80.64%, 100.00%]
Highway Entry	8	0	8	100.00%	[67.56%, 100.00%]
Highway Exit	8	0	8	100.00%	[67.56%, 100.00%]
Living Street	7	1	8	87.50%	[52.91%, 97.76%]
U Turn	7	1	8	87.50%	[52.91%, 97.76%]
Intersection	5,115	72	5,187	98.61%	[98.26%, 98.90%]
Crosswalk	4,200	18	4,218	99.57%	[99.33%, 99.73%]
Traffic Light	3,367	110	3,477	96.84%	[96.20%, 97.37%]
Bus Stop	1,368	0	1,368	100.00%	[99.72%, 100.00%]
Yield Road	728	13	741	98.25%	[97.02%, 98.97%]
Roundabout	321	21	342	93.86%	[90.80%, 95.95%]
Stop Sign	207	21	228	90.79%	[86.33%, 93.90%]
Digital Speed Sign	154	17	171	90.06%	[84.66%, 93.70%]
Avoid Bus Lane	57	0	57	100.00%	[93.69%, 100.00%]
Highway Entry	57	0	57	100.00%	[93.69%, 100.00%]
Yield Bikelane	57	0	57	100.00%	[93.69%, 100.00%]
Highway Exit	40	17	57	70.18%	[57.34%, 80.47%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Digital Speed Sign	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Road Narrowing	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Bikelane	0	0	0	nan%	[0.00%, 0.00%]
Yield Road	0	0	0	nan%	[0.00%, 0.00%]
Yield Sidewalk	0	0	0	nan%	[0.00%, 0.00%]
Intersection	386	4	390	98.97%	[97.39%, 99.60%]
Traffic Light	134	1	135	99.26%	[95.92%, 99.87%]
Priority	95	1	96	98.96%	[94.33%, 99.82%]
Yield Sidewalk	78	0	78	100.00%	[95.31%, 100.00%]
Bus Stop	51	0	51	100.00%	[93.00%, 100.00%]
Digital sign	31	2	33	93.94%	[80.39%, 98.32%]
Crosswalk	24	0	24	100.00%	[86.20%, 100.00%]
Roundabout	18	0	18	100.00%	[82.41%, 100.00%]
Road Narrowing	9	0	9	100.00%	[70.09%, 100.00%]
Stop Sign	6	0	6	100.00%	[60.97%, 100.00%]
Highway Entry	3	0	3	100.00%	[43.85%, 100.00%]
Highway Exit	3	0	3	100.00%	[43.85%, 100.00%]
Living Street	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	384	3	387	99.22%	[97.75%, 99.74%]
Intersection	159	0	159	100.00%	[97.64%, 100.00%]
Yield Road	124	2	126	98.41%	[94.40%, 99.56%]
Traffic Light	113	4	117	96.58%	[91.54%, 98.66%]
Roundabout	86	1	87	98.85%	[93.77%, 99.80%]
Bus Stop	78	0	78	100.00%	[95.31%, 100.00%]
Stop Sign	27	0	27	100.00%	[87.54%, 100.00%]
Yield Bikelane	18	0	18	100.00%	[82.41%, 100.00%]
Road Narrowing	12	0	12	100.00%	[75.75%, 100.00%]
Bridge	9	0	9	100.00%	[70.09%, 100.00%]
Priority Sign	9	0	9	100.00%	[70.09%, 100.00%]
School Zone	9	0	9	100.00%	[70.09%, 100.00%]
Yield Sidewalk	9	0	9	100.00%	[70.09%, 100.00%]
Highway Entry	6	0	6	100.00%	[60.97%, 100.00%]
Avoid Bus Lane	5	1	6	83.33%	[43.65%, 96.99%]
Highway Exit	5	1	6	83.33%	[43.65%, 96.99%]
Yield Bus Lane	5	1	6	83.33%	[43.65%, 96.99%]
Intersection	3,536	1	3,537	99.97%	[99.84%, 100.00%]
Crosswalk	2,024	1	2,025	99.95%	[99.72%, 99.99%]
Yield Road	1,753	2	1,755	99.89%	[99.59%, 99.97%]
Bus Stop	1,377	0	1,377	100.00%	[99.72%, 100.00%]
Traffic Light	1,288	8	1,296	99.38%	[98.79%, 99.69%]
Stop Sign	136	26	162	83.95%	[77.52%, 88.81%]
Roundabout	81	0	81	100.00%	[95.47%, 100.00%]
Avoid Bus Lane	63	18	81	77.78%	[67.58%, 85.46%]
Bridge	27	0	27	100.00%	[87.54%, 100.00%]
Highway Entry	27	0	27	100.00%	[87.54%, 100.00%]
Highway Exit	27	0	27	100.00%	[87.54%, 100.00%]

Interaction Type	Tests Passed	Tests Failed	Total Tests	Pass Rate (%)	95% CI
Intersection	7,585	85	7,670	98.89%	[98.63%, 99.10%]
Crosswalk	2,955	13	2,968	99.56%	[99.25%, 99.74%]
Yield Bikelane	2,113	6	2,119	99.72%	[99.38%, 99.87%]
Yield Road	1,467	3	1,470	99.80%	[99.40%, 99.93%]
Digital Speed Sign	1,103	1	1,104	99.91%	[99.49%, 99.98%]
Roundabout	643	3	646	99.54%	[98.64%, 99.84%]
Yield Sidewalk	618	0	618	100.00%	[99.38%, 100.00%]
Bus Stop	614	2	616	99.68%	[98.82%, 99.91%]
Road Narrowing	607	6	613	99.02%	[97.88%, 99.55%]
Yield Tram	481	0	481	100.00%	[99.21%, 100.00%]
Yield Bus Lane	341	1	342	99.71%	[98.36%, 99.95%]
Bridge	336	7	343	97.96%	[95.85%, 99.01%]
Traffic Light	306	23	329	93.01%	[89.73%, 95.30%]
Avoid Bus Lane	204	1	205	99.51%	[97.29%, 99.91%]
Priority Sign	203	2	205	99.02%	[96.51%, 99.73%]
School Zone	137	0	137	100.00%	[97.27%, 100.00%]
Dynamic Barrier	135	2	137	98.54%	[94.83%, 99.60%]
Highway Entry	69	0	69	100.00%	[94.73%, 100.00%]
U Turn	69	0	69	100.00%	[94.73%, 100.00%]
Highway Exit	68	1	69	98.55%	[92.24%, 99.74%]
Living Street	68	1	69	98.55%	[92.24%, 99.74%]
Intersection	983	18	1,001	98.20%	[97.18%, 98.86%]
Crosswalk	806	8	814	99.02%	[98.07%, 99.50%]
Traffic Light	646	25	671	96.27%	[94.56%, 97.46%]
Bus Stop	263	1	264	99.62%	[97.89%, 99.93%]
Yield Road	142	1	143	99.30%	[96.15%, 99.88%]
Roundabout	66	0	66	100.00%	[94.50%, 100.00%]
Stop Sign	41	3	44	93.18%	[81.77%, 97.65%]
Digital Speed Sign	33	0	33	100.00%	[89.57%, 100.00%]
Avoid Bus Lane	11	0	11	100.00%	[74.12%, 100.00%]
Highway Entry	11	0	11	100.00%	[74.12%, 100.00%]
Yield Bikelane	11	0	11	100.00%	[74.12%, 100.00%]
Highway Exit	7	4	11	63.64%	[35.38%, 84.83%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Digital Speed Sign	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Road Narrowing	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Bikelane	0	0	0	nan%	[0.00%, 0.00%]
Yield Road	0	0	0	nan%	[0.00%, 0.00%]
Yield Sidewalk	0	0	0	nan%	[0.00%, 0.00%]
Bus Stop	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	0	0	0	nan%	[0.00%, 0.00%]
Digital sign	0	0	0	nan%	[0.00%, 0.00%]
Highway Entry	0	0	0	nan%	[0.00%, 0.00%]
Highway Exit	0	0	0	nan%	[0.00%, 0.00%]
Intersection	0	0	0	nan%	[0.00%, 0.00%]
Living Street	0	0	0	nan%	[0.00%, 0.00%]
Priority	0	0	0	nan%	[0.00%, 0.00%]
Road Narrowing	0	0	0	nan%	[0.00%, 0.00%]
Roundabout	0	0	0	nan%	[0.00%, 0.00%]
Stop Sign	0	0	0	nan%	[0.00%, 0.00%]
Traffic Light	0	0	0	nan%	[0.00%, 0.00%]
Yield Sidewalk	0	0	0	nan%	[0.00%, 0.00%]
Crosswalk	644	1	645	99.84%	[99.13%, 99.97%]
Intersection	265	0	265	100.00%	[98.57%, 100.00%]
Yield Road	210	0	210	100.00%	[98.20%, 100.00%]
Traffic Light	190	5	195	97.44%	[94.14%, 98.90%]
Roundabout	145	0	145	100.00%	[97.42%, 100.00%]
Bus Stop	129	1	130	99.23%	[95.77%, 99.86%]
Stop Sign	45	0	45	100.00%	[92.13%, 100.00%]
Yield Bikelane	30	0	30	100.00%	[88.65%, 100.00%]
Road Narrowing	20	0	20	100.00%	[83.89%, 100.00%]
Bridge	15	0	15	100.00%	[79.61%, 100.00%]
Priority Sign	15	0	15	100.00%	[79.61%, 100.00%]
School Zone	15	0	15	100.00%	[79.61%, 100.00%]
Yield Sidewalk	15	0	15	100.00%	[79.61%, 100.00%]
Highway Entry	10	0	10	100.00%	[72.25%, 100.00%]
Highway Exit	10	0	10	100.00%	[72.25%, 100.00%]
Avoid Bus Lane	9	1	10	90.00%	[59.58%, 98.21%]
Yield Bus Lane	9	1	10	90.00%	[59.58%, 98.21%]
Intersection	262	0	262	100.00%	[98.55%, 100.00%]
Crosswalk	150	0	150	100.00%	[97.50%, 100.00%]
Yield Road	130	0	130	100.00%	[97.13%, 100.00%]
Bus Stop	102	0	102	100.00%	[96.37%, 100.00%]
Traffic Light	96	0	96	100.00%	[96.15%, 100.00%]
Stop Sign	7	5	12	58.33%	[31.95%, 80.67%]
Roundabout	6	0	6	100.00%	[60.97%, 100.00%]
Avoid Bus Lane	5	1	6	83.33%	[43.65%, 96.99%]
Bridge	2	0	2	100.00%	[34.24%, 100.00%]
Highway Entry	2	0	2	100.00%	[34.24%, 100.00%]
Highway Exit	2	0	2	100.00%	[34.24%, 100.00%]

Article 39 - FSD (Supervised) Dashboard

Table of Contents

Incompatibility with R171.01

Summary of Requested Exemptions

Evidence Map: Studies Supporting Each Exemption

Overall System Safety Evidence

Full Self-Driving (Supervised) as a Product

Full Self-Driving (Supervised) as an SAE Level 2 System

The Safety Pillars of FSD (Supervised)

Video Clip Evidence of Behavioral and Safety Competency

Control Types

Active Safety Features

Road Classification

SAE Level 2 Definition

Statistical Methods (IRR, Odds Ratios)

Study Takeaways

Part 1 Key Takeaways:

Mileage Breakdown Data

Part 2 Key Takeaways:

Part 3 Key Takeaways:

Part 4 Key Takeaways:

Mileage Breakdown Data

Evidence-Based Rationale for Surrogate Safety Metrics

Study Takeaways

IRR Comparison - AEB Rates by Control Type and Road Class (North America, 2025)

Chart Data

Regression Results

IRR Comparison - AEB Rates by Control Type and Road Class (Europe, Engineering Fleet)

Chart Data

Regression Results

IRR Comparison - Positive Longitudinal Acceleration Events (≥ 3 m/s²)

Rate Data

Statistical Comparisons (FSD vs Manual)

IRR Comparison - Negative Longitudinal Acceleration Events (≥ 3 m/s²) - Braking

Rate Data

Statistical Comparisons (FSD vs Manual)

IRR Comparison - Lateral Acceleration Events (≥ 3 m/s²)

Rate Data

Statistical Comparisons (FSD vs Manual)

IRR Comparison - Horn Usage

Rate Data

Statistical Comparisons (FSD vs Manual)

IRR Comparison - Blinker Usage

Rate Data

Statistical Comparisons (FSD vs Manual)

Study Takeaways

Part 1 Key Takeaway

Statistical Comparisons (FSD vs Manual)

Part 2 Key Takeaway

Statistical Comparisons (FSD vs Manual)

Part 3 Key Takeaway

Manual Driving Baseline: Distribution of Visual Attention Patterns

Manual Driving Baseline: Cumulative Relevant Awareness

FSD (Supervised): Distribution of Visual Attention Patterns

FSD (Supervised): Cumulative Relevant Awareness

Part 4 Key Takeaways

Study Takeaways

Driver Fatigue Data

Statistical Comparison (FSD vs Manual)

Collision Definition

Collision Classification

Data Period

Study Takeaways

Why Test for Overdispersion by Control Type and Road Class?

Per-Group Overdispersion Test Results

Model Selection Based on Overdispersion

IRR Comparison - Major Incidents per Million Miles (2025, North America)

Part 1 Key Takeaways

Chart Data

Regression Results

IRR Comparison - Major Incidents - Year over Year Comparison (2024 vs 2025, North America)

Part 2 Key Takeaways

Year-over-Year Change Summary

Regression Results

Manual (No Active Safety Features)

Manual (With Active Safety Features)

FSD (Supervised, HW4)

Part 3 Key Takeaways

IRR Comparison - Minor Incidents per Million Miles (2025, North America)

Part 4 Key Takeaways