Question 1

How many autonomous vehicle crashes have been reported?

Accepted Answer

As of April 2026, NHTSA's Standing General Order database contains 6,215 incidents involving automated driving systems. Of these, 2,609 involve fully autonomous ADS vehicles (like Waymo robotaxis) and 3,606 involve driver-assist ADAS systems (like Tesla Autopilot).

Question 2

How many people have died in autonomous vehicle crashes?

Accepted Answer

The database records 68 fatalities. Tesla accounts for 56 of those deaths (82%). Waymo and General Motors each have 2. The remaining fatalities are spread across other manufacturers including Ford, Honda, and Nissan.

Question 3

Which company has the most AV incidents?

Accepted Answer

Tesla leads with 3,092 reported incidents — roughly half the entire database. This reflects Tesla's enormous consumer fleet (millions of vehicles with Autopilot/FSD) compared to robotaxi operators with thousands of vehicles. Waymo is second with 1,729 incidents but only 2 fatalities.

Question 4

Is Tesla Autopilot safe?

Accepted Answer

The data paints a mixed picture. Tesla's 3,092 incidents include 56 fatalities — far more than any other manufacturer. While Tesla's large fleet partially explains the high incident count, the fatality rate remains concerning. NHTSA is actively investigating 3.2 million Tesla vehicles for camera visibility failures in FSD.

Question 5

What is phantom braking?

Accepted Answer

Phantom braking occurs when a vehicle's automated system perceives a non-existent obstacle and initiates sudden, unexpected braking. It's the most common Tesla Autopilot complaint and worsened significantly after Tesla removed radar sensors in 2021, relying solely on cameras ("Tesla Vision").

Question 6

Where do most AV crashes happen?

Accepted Answer

California leads with 2,739 incidents (44% of the national total), driven largely by Waymo's San Francisco robotaxi operations. Texas is second (684), followed by Arizona (567), Florida (303), and Georgia (151). San Francisco alone accounts for 1,170 incidents.

Question 7

What time of day do most AV crashes occur?

Accepted Answer

Crashes peak around 4:00 PM (16:00), coinciding with afternoon rush hour. There's a secondary peak during the morning commute (8–9 AM). Overnight hours (1–5 AM) see the fewest incidents but proportionally higher severity.

Question 8

How fast are cars going when they crash?

Accepted Answer

The average pre-crash speed across all reported incidents is approximately 36 mph. However, ADAS incidents (like Tesla Autopilot crashes) tend to occur at higher speeds (often highway speeds of 55+ mph), while ADS robotaxi incidents typically happen at lower urban speeds (under 25 mph).

Question 9

What is the difference between ADS and ADAS?

Accepted Answer

ADS (Automated Driving System) means the vehicle drives itself with no human intervention required within its operating domain — like a Waymo robotaxi (Level 3–5). ADAS (Advanced Driver Assistance System) means a human driver is required but gets assistance — like Tesla Autopilot or GM Super Cruise (Level 1–2).

Question 10

Can I check if my car has been recalled?

Accepted Answer

Yes! Use our Recall Checker tool to search by manufacturer, model, or VIN. We track all NHTSA recalls related to autonomous and driver-assist systems, including Tesla's OTA software recalls affecting millions of vehicles.

Question 11

Are autonomous vehicles safer than human drivers?

Accepted Answer

It's complicated. Fully autonomous ADS systems (like Waymo) show promising safety records in controlled urban environments — 1,729 incidents but only 2 fatalities. ADAS systems (like Tesla) have a worse fatality profile: 3,092 incidents and 56 deaths. Human drivers cause ~40,000 deaths per year across billions more miles driven. Direct comparison requires careful normalization by VMT, which most AV companies don't fully disclose.

Question 12

Why does Tesla have so many more incidents than other companies?

Accepted Answer

Three factors: (1) Fleet size — Tesla has millions of ADAS-equipped vehicles vs. thousands for robotaxi companies; (2) Reporting scope — SGO requires reporting for all Level 2+ crashes, and Tesla's large fleet generates more reports; (3) Genuine safety concerns — Tesla's camera-only approach and aggressive feature naming may contribute to driver over-reliance.

Question 13

Which vehicle model has the most incidents?

Accepted Answer

The Jaguar I-PACE (used by Waymo's robotaxi fleet) leads with 1,763 incidents, followed by the Tesla Model Y (1,295), Tesla Model 3 (1,289), Cruise AV (293), and Tesla Model X (251).

Question 14

What is NHTSA's Standing General Order (SGO)?

Accepted Answer

SGO 2021-01 is a federal regulation issued June 29, 2021 requiring manufacturers to report crashes involving their automated driving systems. Fatal or serious-injury crashes must be reported within 1 day; other qualifying crashes within 10 days. This regulation created the dataset that makes AutoPilotWatch possible.

Question 15

Where does this data come from?

Accepted Answer

All incident data comes from NHTSA's Standing General Order 2021-01 mandatory crash reporting program. We supplement this with NHTSA recall data, investigation records, Technical Service Bulletins (TSBs), consumer complaints, and California DMV autonomous vehicle reports. All sources are public government data.

Question 16

How often is the data updated?

Accepted Answer

AutoPilotWatch processes the latest NHTSA SGO data with each release. The database was last updated on April 15, 2026. We aim for monthly refreshes as NHTSA publishes updated data. Subscribe to our newsletter for update notifications.

Question 17

Can I download the raw data?

Accepted Answer

Yes! Visit our Downloads page for processed JSON data files, CSV exports, and direct links to original NHTSA sources. All data is free and open for research, journalism, and public interest use.

Question 18

Why does Tesla use cameras only while others use lidar?

Accepted Answer

Tesla CEO Elon Musk has argued that cameras alone can replicate human vision at lower cost. Starting in 2021, Tesla removed radar and ultrasonic sensors, relying solely on cameras ("Tesla Vision"). Every other major AV company uses multi-sensor stacks including lidar and radar for redundancy. NHTSA's current investigation specifically examines camera visibility failures in glare, fog, and dust.

Question 19

What happened to Cruise?

Accepted Answer

Cruise LLC, GM's robotaxi subsidiary, accumulated 155 incidents before halting operations in late 2023 after a pedestrian-dragging incident in San Francisco. The California DMV revoked Cruise's driverless testing permit, and GM subsequently wound down Cruise as a standalone robotaxi operation.

Question 20

How many manufacturers report AV crash data?

Accepted Answer

Currently 60 manufacturers have reported incidents to NHTSA under the SGO. The top reporters are Tesla (3,092), Waymo (1,729), General Motors (265), Cruise (155), Transdev (140), and Zoox (120).

Question 21

What is an Operational Design Domain (ODD)?

Accepted Answer

The ODD defines the specific conditions where an automated system is designed to work — including geography, road types, speed ranges, weather, and time of day. A robotaxi approved for sunny Phoenix streets cannot necessarily operate in snowy Minneapolis highways. Operating outside the ODD is a major safety risk.

Question 22

How does AutoPilotWatch calculate safety scores?

Accepted Answer

Our Safety Scorecard uses a weighted formula considering: incident rate per reported vehicle, fatality count, injury severity, recall responsiveness, and investigation history. Scores are relative rankings — not absolute safety certifications. See our full methodology for details.

Question 23

Is this site affiliated with NHTSA or any car manufacturer?

Accepted Answer

No. AutoPilotWatch is an independent project of The Data Project (TheDataProject.ai). We are not affiliated with NHTSA, Tesla, Waymo, or any vehicle manufacturer. We process publicly available government data to make it searchable and understandable.

Question 24

I found an error in the data. How do I report it?

Accepted Answer

We take accuracy seriously. Please email info@thedataproject.ai with the specific page URL and what needs correction, or use our contact form. We investigate all reports and issue corrections promptly.

Question 25

What is a Standing General Order (SGO) report?

Accepted Answer

An SGO report is a mandatory crash filing required by NHTSA under Standing General Order 2021-01. Manufacturers must report qualifying crashes involving their automated driving systems within strict timelines — 1 day for fatal/serious crashes, 10 days for others. These reports form the foundation of AutoPilotWatch's incident database.

Question 26

What is the difference between a recall and a TSB?

Accepted Answer

A recall is a mandatory safety action where manufacturers must notify all affected owners and provide a free fix. A Technical Service Bulletin (TSB) is optional guidance sent to dealers describing known issues — owners may not be notified. TSBs often precede recalls and can signal emerging safety patterns.

Question 27

How does AutoPilotWatch handle confidential business information (CBI)?

Accepted Answer

Some fields in NHTSA's SGO reports are redacted by manufacturers as Confidential Business Information. This may include software version details, specific system configurations, and narrative descriptions. We display all publicly available fields and note when data has been redacted.

Question 28

What is sensor fusion and why does it matter?

Accepted Answer

Sensor fusion combines data from multiple sensor types — cameras, lidar, radar, and ultrasonic — to create a more reliable understanding of the driving environment. Most AV companies use sensor fusion for redundancy. Tesla's controversial camera-only approach forgoes this redundancy, which NHTSA is investigating in connection with visibility-related crashes.

Question 29

Can I embed AutoPilotWatch data or charts on my website?

Accepted Answer

Our data is free to use with attribution. For embedding charts, we recommend downloading the JSON data files and creating your own visualizations. If you'd like to discuss data partnerships or API access, contact us at info@thedataproject.ai.

Question 30

What is Vehicle Miles Traveled (VMT) and why is it important?

Accepted Answer

VMT measures total miles driven by a fleet. It's the gold standard for normalizing crash rates — without it, you can't fairly compare manufacturers with different fleet sizes. A company with 10 crashes per billion miles is safer than one with 5 crashes per million miles. Most AV companies don't publicly disclose VMT, which limits our ability to calculate true safety rates.

Question 31

Does AutoPilotWatch track international AV incidents?

Accepted Answer

Currently, AutoPilotWatch focuses exclusively on U.S. incidents reported to NHTSA. International AV incident data (EU, UK, China, Japan) is on our roadmap but not yet integrated. Different countries have different reporting requirements, making cross-border comparisons challenging.

Question 32

What happens when NHTSA opens an investigation?

Accepted Answer

NHTSA's Office of Defects Investigation (ODI) opens investigations when it identifies a potential safety defect pattern. Investigations can be triggered by crash data, consumer complaints, or media reports. If the investigation confirms a defect, it can lead to a recall. NHTSA is currently investigating Tesla's FSD system under case EA26-001.

Question 33

How do over-the-air (OTA) updates affect safety data?

Accepted Answer

OTA updates can fix software defects that caused safety issues — but they also change the system being studied. A phantom braking fix deployed via OTA may reduce future incidents, but the pre-fix incidents remain in the database. We track software version data where available to help researchers correlate updates with safety outcomes.

Frequently Asked Questions

💡 Quick Overview

Data Coverage

Independence

Free & Open

About These FAQs

Explore the Data