Curated Snapshot - Model Guide
Before You Buy
Every year the government logs roughly a thousand vehicle recalls covering tens of millions of cars. Most are routine. A handful were lethal, and a few of those took the manufacturer years to admit. This is a plain reading of the recalls that actually mattered - which defects killed people, which ones dragged, and how to read a crash-test rating before you sign anything.
Most Recalled
I. Landmark Campaigns by Vehicles AffectedThe seven campaigns below are the ones a used-car shopper should actually know by name. Ranked by how many vehicles (or, for Takata, inflators) were pulled back. Takata's 42M dwarf the rest, so the bars use a log scale - each gridline is 10x the last (1M, 10M, 100M) - which keeps a 1.6M recall legible without pretending it rivals Takata. The exact count sits at the end of every row. The orange bars and the flag mark defects tied to confirmed deaths; a Do Not Drive order means exactly that.
- 01
Takata airbag inflators
Do Not DriveAmmonium-nitrate inflators degrade with age and humidity, then rupture - firing metal shrapnel into the cabin.
42M42M+ vehicles / 67M inflators - 02
Toyota unintended acceleration
Accelerator pedals could be trapped by the floor mat or stick from assembly friction.
8Mabout 8M vehicles - 03
Ford / Firestone tires
Tread separated from the steel belt at speed, sending the Ford Explorer into a rollover.
6.50M6.5M tires recalled (14.4M produced) - 04
GM ignition switch
The switch could slip out of 'run' if jostled, cutting the engine, power steering and airbags.
2.60M2.6M vehicles - 05
Tesla Autopilot (Autosteer)
Autosteer's controls did not do enough to keep drivers engaged, allowing foreseeable misuse.
2M2M+ vehicles - 06
Hyundai / Kia Theta II engines
Machining debris could restrict oil flow, seize the connecting-rod bearing and start a non-crash fire.
1.60M1.6M+ vehicles - 07
Jeep rear fuel tank
A plastic fuel tank mounted behind the rear axle could rupture and ignite in a rear-end crash.
1.56M1.56M vehicles
All seven campaigns, in numbers
| # | Campaign | Units | U.S. deaths | Recalled |
|---|---|---|---|---|
| 1 | Takata airbag inflators | 42M+ vehicles / 67M inflators | 28 | 2014-2020, ongoing |
| 2 | Toyota unintended acceleration | about 8M vehicles | - | 2009-2010 |
| 3 | Ford / Firestone tires | 6.5M tires recalled (14.4M produced) | 238 | 2000 |
| 4 | GM ignition switch | 2.6M vehicles | 124 | 2014 |
| 5 | Tesla Autopilot (Autosteer) | 2M+ vehicles | - | 2023 |
| 6 | Hyundai / Kia Theta II engines | 1.6M+ vehicles | - | 2015, expanded 2017 |
| 7 | Jeep rear fuel tank | 1.56M vehicles | 44 | 2013 |
The Slow Recalls
II. First Warning to Recall IssuedA recall is only as fast as the manufacturer allows. Each line runs from the first documented warning sign to the day the recall was finally issued. The open ring is the warning; the filled dot is the recall. The distance between them is measured in years - and, too often, in deaths.
The timelines, in numbers
| Campaign | First warning | Recalled | Gap | Deaths |
|---|---|---|---|---|
| Takata airbag inflators | 2004 (First known inflator rupture, on a Honda in Alabama.) | 2015 | 11 yr | 28 |
| GM ignition switch | 2004 (GM engineers documented the switch slipping during development.) | 2014 | 10 yr | 124 |
| Hyundai / Kia Theta II | 2011 (Earliest engine-fire reports on 2011 Sonatas.) | 2015 | 4 yr | - |
| Jeep rear fuel tank | 2010 (NHTSA opened a defect investigation into rear-impact fires.) | 2013 | 3 yr | 44 |
| Ford / Firestone tires | 1998 (State Farm flagged a pattern of tread-separation claims to NHTSA.) | 2000 | 2 yr | 238 |
Reading a Crash Test
III. What the Stars Do and Don't SayNHTSA's New Car Assessment Program is the source of the five-star rating on the window sticker. It is genuinely useful - and routinely misread. Here is what each rating actually measures, and two things it quietly leaves out.
Head-on into a fixed barrier at 35 mph, with belted dummies in the two front seats.
A moving barrier plus a rigid pole strike the driver's side.
A static tip measure plus a dynamic maneuver - the only non-crash rating.
A single 1-to-5-star summary, added when NHTSA overhauled the program in 2011.
NHTSA's New Car Assessment Program rates crash protection from one to five stars. Two cautions for a buyer: stars are graded within a weight class, so a five-star small car is not a five-star large car; and the frontal test does not include a small-overlap crash, which the separate IIHS program added in 2012 and on which several five-star cars initially performed poorly. The card at right shows one real vehicle's ratings, read live from the SafetyRatings API; every model's ratings are searchable at nhtsa.gov.
Complaint Hotspots
IV. Where Owners File - by ComponentBefore a defect becomes a recall it usually shows up first as a pile of owner complaints in NHTSA's public database, sorted into standard component families. Counted across every complaint NHTSA has logged since 1995 - about 2.2 million of them - these are the families that draw the most reports.
Real complaint counts, NHTSA FLAT_CMPL · % is each family's share of all 2,224,108 logged complaints. The eight families shown cover 66% of the total.
Methodology
V. What Is Real, What Is IllustrativeVehicle-safety data affects real purchase decisions, so the honesty rule here is strict: no vehicle-specific recall, death, or crash-test number is invented. Everything presented as fact is a well-documented public record or a direct count from NHTSA's own bulk files. Anything that names a live source instead of asserting a figure is labelled live data slot right where it appears.
Source
The data is built from NHTSA's own Datasets & APIs: the recalls flat files (FLAT_RCL, 30,143 distinct campaigns across 324,369 rows), the complaints flat file (FLAT_CMPL, 2,224,108 logged complaints since 1995), and the NCAP crash-test SafetyRatings API - https://www.nhtsa.gov/nhtsa-datasets-and-apis. The fetch recipe and swap-point (scripts/fetch-nhtsa.sh, src/lib/source.ts, scripts/build-data.ts) are documented in HANDOFF.md; re-running them refreshes the figures with the page unchanged.
What is real
The Complaint Hotspots bars are now direct tallies from the 2,224,108-row complaints file, counted per component family (NHTSA's own taxonomy has changed over the years, so the historical variants of a family - for example the legacy "Engine" and later "Engine and Engine Cooling" - are folded together; the counts are real). The Reading a Crash Test card shows a real vehicle's real ratings, read from the SafetyRatings API. The seven landmark campaigns, their death tolls and the Slow Recalls gaps come from NHTSA records, consent orders, victim-compensation funds and contemporaneous reporting; the "largest recall in U.S. history" figure is Takata's documented 67M-inflator total. The NCAP section describes the real program and two genuine limitations.
What names a source rather than asserting a figure
On the per-vehicle cards in Compare, the NCAP star ratings and per-vehicle complaint totals are shown as live data slots: they name the free NHTSA source they join from (SafetyRatings, FLAT_CMPL) rather than printing a number. That is deliberate - each landmark vehicle spans many model years, and a crash rating is specific to one model-year-and-trim, so a single star figure would misrepresent the span. This is the one surface still marked pending, which is why the snapshot is labelled curated rather than full.
On the timelines
The first-warning years in Slow Recalls are documented-approximate. A defect rarely has a single clean "first warning" date; we use the earliest widely-reported signal - an internal engineering note, a regulator's investigation, an insurer flagging a claims pattern - and pair it with the recall's issue year. Read the gaps as honest orders of magnitude, not to-the-day precision.
Recall severity and "Do Not Drive"
Not all recalls are equal. Most are routine fixes. A Do Not Drive advisory - which NHTSA and a manufacturer issue jointly - means the risk of continuing to drive outweighs the inconvenience of parking the car until it is repaired. On this page it is reserved for exactly that: as of writing, certain unrepaired 2001-2003 Honda/Acura Takata "Alpha" inflators. Recall repairs are free; check any VIN at nhtsa.gov/recalls.
Generated 2026-07-11 13:37 UTC
Source: NHTSA recalls flat files (FLAT_RCL PRE/POST 2010) + complaints (FLAT_CMPL) + SafetyRatings API; landmark campaigns curated from public records