7,713 crash reports name a driver-assist system. Not one has been verified.
Crashes with a driver-assist system engaged, as the manufacturers reported them
7,713 reports filed against 370 model years on this site, Aug 2019 to Sep 2026: 5,014 with a Level 2 system (lane centering plus adaptive cruise) engaged and 2,659 with an automated-driving system. Each one is a notice the manufacturer or operator was ordered to file within days of hearing of a crash — before verifying it, and only if it heard of it at all. Read the method before the tables.
What these numbers are, and are not
Source: NHTSA's own data-limitations notes, restatedA report is a notice, not a finding. The order requires the reporting entity to file within five days of learning of a serious crash and monthly for the rest, whether or not it has verified anything. Updates routinely change the story: several on file record that the event data showed no system was engaged after all. Nothing here has been verified by NHTSA or by us.
Counts measure the reporting as much as the car. A manufacturer with vehicle telematics learns of an air-bag deployment within minutes; one without it learns of a crash only if the owner complains and mentions the system. NHTSA says plainly that a higher count can mean better data access rather than more crashes. Do not rank makes by this page.
Nothing is normalised. Entities do not report how many equipped vehicles they have on the road or how far they travel, so there is no rate here — only counts.
What is reportable differs. A Level 2 crash is reportable only with a fatality, a hospital transport, an air-bag deployment or a vulnerable road user struck; an automated-driving crash has a lower property-damage threshold. The two columns are not the same measure.
What the page can tell you: which model years on this site carry reports, what system was named, what was struck, how the injuries were alleged, and where to read the filing. Every count links the vehicle page, which quotes the manufacturer’s narrative where NHTSA published it. Source: NHTSA, Standing General Order on Crash Reporting, data files reposted monthly.
Model years with reports on file
Source: reports joined to this site's pages| Model year | Reports | L2 / ADS | System named | Struck | Fatal | Injury alleged | Filed for |
|---|---|---|---|---|---|---|---|
| 2024 Jaguar I-Pace | 1725 | 1 / 1724 | 5th Generation ADS, Version 10, 5th Generation ADS | Passenger Car, SUV | 3 | 46 | Feb 2024 – Sep 2026 |
| 2021 Jaguar I-Pace | 540 | 1 / 533 | 5th Generation ADS, 5th Generation | Passenger Car, SUV | 0 | 28 | Jul 2021 – Sep 2026 |
| 2026 Tesla Model Y | 462 | 419 / 35 | ADS | Unknown, Other | 3 | 8 | Apr 2025 – Sep 2026 |
| 2023 Tesla Model Y | 440 | 434 / 0 | — | Unknown, Other | 5 | 7 | Jan 2023 – Sep 2026 |
| 2021 Tesla Model Y | 302 | 302 / 0 | — | Other, Unknown | 6 | 8 | Jul 2021 – Sep 2026 |
| 2024 Tesla Model Y | 279 | 277 / 0 | — | Unknown, Other | 3 | 6 | Apr 2024 – Sep 2026 |
| 2025 Tesla Model 3 | 273 | 273 / 0 | — | Unknown, Other | 3 | 5 | Oct 2024 – Sep 2026 |
| 2022 Tesla Model Y | 270 | 267 / 0 | — | Unknown, Other | 4 | 4 | Feb 2022 – Sep 2026 |
| 2023 Tesla Model 3 | 255 | 254 / 0 | — | Unknown, Other | 4 | 6 | Jan 2023 – Sep 2026 |
| 2018 Tesla Model 3 | 241 | 241 / 0 | — | Unknown, Passenger Car | 7 | 7 | Oct 2020 – Sep 2026 |
| 2020 Tesla Model 3 | 198 | 197 / 0 | — | Other, Unknown | 7 | 7 | Jul 2020 – Sep 2026 |
| 2022 Tesla Model 3 | 191 | 191 / 0 | — | Unknown, Other | 4 | 6 | Jan 2022 – Sep 2026 |
Sorted by reports filed; the caveats above apply to every row.
By make
| Make | Reports | Model years | Fatal | Injury alleged |
|---|---|---|---|---|
| Tesla | 4,433 | 42 | 66 | 96 |
| Jaguar | 2,268 | 3 | 3 | 74 |
| Toyota | 212 | 30 | 1 | 27 |
| Hyundai | 134 | 23 | 0 | 3 |
| Honda | 91 | 37 | 1 | 11 |
| Ford | 83 | 25 | 3 | 13 |
| Subaru | 66 | 20 | 1 | 25 |
| Cadillac | 57 | 21 | 1 | 3 |
| Chevrolet | 57 | 12 | 2 | 4 |
| GMC | 54 | 12 | 0 | 1 |
| Nissan | 32 | 13 | 1 | 5 |
| BMW | 31 | 26 | 0 | 8 |
| Kia | 27 | 17 | 1 | 18 |
| Mercedes-Benz | 26 | 17 | 0 | 7 |
| Acura | 25 | 14 | 0 | 0 |
| Rivian | 21 | 6 | 0 | 1 |
| Lexus | 16 | 6 | 0 | 0 |
| Chrysler | 14 | 2 | 0 | 0 |
| Lucid | 11 | 4 | 0 | 0 |
| Lincoln | 10 | 6 | 0 | 0 |
| Jeep | 8 | 5 | 0 | 1 |
| Volkswagen | 7 | 6 | 0 | 0 |
| VinFast | 5 | 2 | 0 | 0 |
| Volvo | 5 | 5 | 0 | 1 |
| Buick | 4 | 2 | 0 | 0 |
| Ram | 3 | 2 | 0 | 0 |
| Audi | 2 | 2 | 0 | 0 |
| Dodge | 2 | 2 | 0 | 1 |
| Polestar | 2 | 1 | 0 | 1 |
| Porsche | 2 | 2 | 0 | 1 |
| Genesis | 1 | 1 | 0 | 0 |
| Infiniti | 1 | 1 | 0 | 0 |
| MINI | 1 | 1 | 0 | 1 |
| Maserati | 1 | 1 | 0 | 0 |
| Mazda | 1 | 1 | 0 | 0 |
Systems named in the filings
As the reporting entity wrote them; about a third are withheld as confidential business information.
What was struck
Crash partner
Highest injury alleged
By month of incident
Source: last 24 months on fileOct 2025 224 · Nov 2025 227 · Dec 2025 238 · Jan 2026 230 · Feb 2026 216 · Mar 2026 292 · Apr 2026 310 · May 2026 334 · Jun 2026 348 · Jul 2026 363 · Aug 2026 316 · Sep 2026 90
Get the data
One row per model year, with the counts on this page, in CSV and JSON (the JSON also carries each report’s month, system, crash partner, severity, state and NHTSA report id). Every vehicle page’s own JSON carries its reports too. Report ids are NHTSA’s; look one up in the agency’s files by that id. Nothing VIN-shaped is in any of these files.
Released under CC BY 4.0 — redistribute it, build on it, sell what you build. Credit WhichTrim and link back. The federal records underneath are public domain; the join, the filtering and the arithmetic are ours. See the carve-out in the terms.
The data behind this
Everything above, as files. No key, no form, no attribution beyond the line below — if it is useful, use it.
How to cite
WhichTrim, “Driver-assist crash reports by model year, from NHTSA's Standing General Order”, 2026-11-04. https://whichtrim.com/driver-assist-crashes/
The underlying records are U.S. federal public-domain data from NHTSA. The joins, the filtering and the arithmetic are ours, and we would rather be credited than not — but the figures are checkable from the source either way, which is the point.