China Just Recalled 4.3 Million Cars Because Drivers Can't Find the Door Release. We Counted Every US Complaint From Someone Who Couldn't Get Out

Key facts

• 4,324 of 1,552,176 NHTSA complaints (0.279%, about 1 in 359) describe an occupant unable to get out of the vehicle.
• 2 complaints in the entire file explicitly say the driver could not find the emergency or manual door release. Both are Teslas.
• China recalled ~4.3 million vehicles across Tesla and at least nine other automakers; Tesla's portion is just under 3 million, its largest ever.
• There is no US recall. NHTSA denied the Model 3 defect petition on 2026-07-24 and instead opened rulemaking. FMVSS 206 says nothing about where a manual release must be or how it must be labelled.
• Tesla's share tripled after the story broke — 0.379% pre-2025, 1.229% in 2025-2026 — while every other top brand's share fell.

What exactly did China recall, and why?

China's market regulator accepted recall filings covering roughly 4.3 million vehicles because the interior emergency mechanical door release is too hard to find. The releases exist; regulators concluded they are coloured close to the surrounding trim and are not identifiable quickly by an occupant or a rescuer, particularly after a crash has cut electrical power.

Tesla filed for just under 3 million cars, the largest recall in the company's history. Xiaomi filed for 390,435 SU7s, Leapmotor for 371,200 vehicles, and Zeekr for 92,658. Chery, Geely, Dongfeng, Arcfox, FAW and Xpeng also filed. Remedies begin in late September 2026.

The trigger was not a statistic. It was a fatal crash in which a driver was trapped in a burning car while bystanders could not get the doors open.

Does the United States have the same problem?

The US has the same hardware and none of the rules. This is the gap that matters for anyone buying a used EV or a modern car with flush handles.

FMVSS 206 exists to stop doors flying open in a crash. Nothing in it requires that a person inside can get a door open afterwards. That is the hole the rulemaking is meant to fill, and rulemaking takes years — so for the used car you are looking at this month, no fix is coming.

Which used vehicles have the most "I couldn't get out" complaints?

Ranked by each brand's egress complaints as a share of that brand's own complaints, restricted to brands with at least 30 such complaints:

At model level, with at least 15 egress complaints:

Did the Tesla ranking survive a publicity check?

No, and that is the most important thing in this dataset. Complaint counts respond to news coverage, so we recomputed every share in a pre-publicity window (complaints received 2015-2024) and a post-publicity window (2025-2026), the period when Bloomberg's reporting landed, NHTSA opened PE25010, and the door-handle story became national news.

Tesla ranks third across all years. In the pre-publicity window it ranks seventh, at 0.379%. In 2025-2026 its share is 1.229% — about 3.2 times higher.

Every other top brand moved the other way. BMW fell from 0.693% to 0.233%. Hyundai fell from 0.853% to 0.368%. Kia fell from 0.400% to 0.194%.

Tesla being the only brand to rise sharply is genuinely ambiguous, and we are not going to resolve it here. It is consistent with a real and growing problem in a fleet that is young and heavily electronic. It is equally consistent with owners who have read the coverage now filing reports that owners of a 2003 sedan never filed. A complaint file cannot separate those two. Anyone citing the all-years ranking without the window split is citing an artefact.

What do these complaints actually describe?

Mostly not futuristic handles. The bulk of the file is ordinary mechanical failure on ordinary cars:

"THE PASSENGER SIDE DOOR JAMS, CAUSING PASSENGERS TO BE TRAPPED IN VEHICLE." — 1995 Mazda 626

"INSIDE HANDLES BREAK OFF, UNABLE TO EXIT VEHICLE." — 1991 Geo Prizm

"POWER DOOR LOCKS INTERMITTENTLY FAIL TO UNLOCK FOR 2-5 MINUTES WHEN IGNITION IS TURNED OFF, OCCUPANTS ARE UNABLE TO EXIT VEHICLE DURING THIS TIME." — 1991 Ford Taurus

The two that name the modern failure read differently:

"THE REAR SEAT OCCUPANTS WERE UNABLE TO OPEN THE FALCON DOORS. THE EMERGENCY DOOR RELEASE IS NOT OBVIOUS TO OCCUPANTS AND REQUIRES THEM TO REACH UP BEHIND THE SPEAKER GRILL TO FIND THE PULL." — 2016 Tesla Model X

The second is a 2024 Model Y that lost all 12-volt power in a parking garage at about 110°F with animals inside.

Two in 1.55 million is not evidence the problem is rare. It is evidence that US owners do not describe it in those words — and that a defect can be serious enough to trigger the largest recall in Chinese history while leaving almost no trace in the US complaint record.

Which of these cars would a recall check actually catch?

This is where a buyer gets caught out, and the two extremes are both in our top three models.

Hyundai Tucson — 3.018%, no recall. The lock-actuator and latch fault behind these complaints was never recalled. Hyundai issued a service bulletin and extended warranty coverage instead. A recall lookup on an affected Tucson comes back clean, because there is nothing to find.

Fisker Ocean — 3.851%, recall exists, remedy does not. Recall 24V466 covers 8,204 US vehicles for door handles that may stick and trap occupants. Fisker is bankrupt. The recall is real, open, and in practice very hard to have performed.

A recall check tells you what a manufacturer was compelled to admit. It does not tell you what a car does.

What can a VIN check tell you here, and what can't it?

It cannot tell you whether you can get out of the car. No vehicle history report can, and we are not going to pretend otherwise. Whether a manual release is findable in the dark is a property of the design and of your hands, not of the vehicle's record.

What a Zilocar VIN check does cover is the risk around it:

• Open safety recalls, including door, latch and handle campaigns — so you know if the car is named in one and whether it is a Fisker-style orphan.
• Accident and damage records, including airbag deployment. A car that has taken a significant side impact is a car whose door wiring, latches and actuators have been disturbed and possibly repaired badly.
• Salvage and junk auction records. Flood and fire history is the single best predictor of future electrical failure, and electronic latches are electrical.
• Odometer readings and rollback flags, theft records, ownership and sales-listing history.

What it does not cover, and what you should get from NHTSA or a dealer instead: whether a recall has actually been remedied on that specific car, the status of open investigations like PE25010, and per-unit firmware.

The check to do yourself, in the car, in 60 seconds

1. Open every door from the inside — including both rears — with the ignition off and the car locked.
2. Find the manual release on each door without searching for it first. If you cannot find it sitting calmly in daylight, you will not find it inverted in the dark. On many EVs the rear manual release is under a mat, behind a panel or beneath the seat, and on some cars the rear doors have no obvious release at all.
3. Check the child locks are off on both rear doors, then re-test from inside.
4. Ask what happens when the 12-volt battery dies. On PE25010's population, low voltage is precisely the failure mode.
5. Sit in the back and try it. Nearly every fatal case in the reporting involved rear-seat occupants.

Method

Source. NHTSA Office of Defects Investigation consumer complaints flat file, FLAT_CMPL.txt, extracted from FLAT_CMPL.zip, retrieved 2026-08-21 and processed 2026-08-23. The file covers complaints received since 1 January 1995.

Universe. 2,237,179 records read. Complaints were deduplicated on ODINO, because a single complaint is repeated once per component, and restricted to records carrying a make and a plausible vehicle model year of 1990-2027. That leaves N = 1,552,176 unique complaints, the denominator for every share on this page.

Classification. The free-text CDESCR field (position 20 of the CMPL record layout) was matched against five explicit rules: (A) a door will not open from the inside; (B) an occupant is trapped or locked in; (C) someone could not exit or get out of the vehicle; (D) a window was broken in order to get out; (E) the emergency or manual door release could not be found. Counts by rule: A 1,439, B 1,474, C 1,390, D 19, E 2. Total 4,324.

Cleanup. Narratives mentioning a seat belt, lap or shoulder belt, buckle, harness or child restraint are excluded outright — restraint entrapment is a real hazard but a different defect, and it was the largest false-positive class in an earlier draft of this analysis. Also excluded: trunks, liftgates, tailgates, hatches, gloveboxes, fuel-filler doors, cargo areas, garage doors, doors that will not stay open, "out of park", and non-occupant uses of "trapped" such as debris in a blower motor.

Validation. The classifier is pinned by a 40-case test suite built from real narratives drawn from this file — 17 that must match and 23 that must not, including every false positive found by reading samples. Matched narratives were read by category before publication. Rules A, B and C, which carry 99.5% of the hits, were clean on inspection; rule D (n=19) retains an estimated four crash-glass false positives and is too small to affect any share reported here.

Aggregation. A brand's share is its egress complaints divided by its own total complaints in the same window. Brands are ranked only with at least 30 egress complaints, models only with at least 15 and 400 total complaints; below that, single reports swing a share by whole percentage points. Every share is recomputed for complaints received 2015-2024 and 2025-2026 to expose publicity effects.

Limits, stated plainly. Complaints are voluntary and unverified. This file has no denominator of vehicles on the road, so nothing here is a failure rate — these are shares of complaints, a measure of composition, not risk. Complaint volume tracks sales volume, vehicle age and press coverage. The VIN field is truncated to 11 characters, so complaints cannot be tied to a specific car. Older records are terse and newer ones verbose, which affects any text-matching method. And an unknown number of genuine egress events are described in words none of our five rules capture; the counts here are a floor, not a census.

Sources

• NHTSA ODI consumer complaints flat file — https://static.nhtsa.gov/odi/ffdd/cmpl/FLAT\_CMPL.zip — Last verified 2026-08-23
• Frozen dataset for this analysis (CSV) — https://zilocar-news.github.io/zilocar-rss/data/used-cars-occupant-egress-complaints-nhtsa-data-2026-08-23.csv — Last verified 2026-08-23
• NHTSA recall lookup — https://www.nhtsa.gov/recalls — Last verified 2026-08-23
• China SAMR recall reporting, Reuters / Bloomberg / CnEVPost, 21 August 2026 — Last verified 2026-08-23
• NHTSA denial of Model 3 defect petition and grant of egress rulemaking, 24 July 2026 — Last verified 2026-08-23
• NHTSA PE25010, Tesla Model Y door handles, opened 15 September 2025 — Last verified 2026-08-23
• Fisker recall 24V466, door handles — https://static.nhtsa.gov/odi/rcl/2024/RCONL-24V466-4996.pdf — Last verified 2026-08-23