Are Used EVs a Bigger Fire Risk Than Gas Cars? We Split Every 'Park Outside' Recall by Powertrain

Key facts

• Dataset: NHTSA Recalls flat file FLAT_RCL_POST_2010.txt, retrieved 2026-08-31, 244,925 rows. Universe = vehicle recall campaigns from 2015 onward flagged PARK_OUTSIDE or matching park-outside/fire keyword language: 113 campaigns, 13,666,010 potentially-affected units.
• By campaign count: Gas/diesel (ICE) only 59 (52.2%); battery-electric (BEV) only 35 (31.0%); mixed ICE + electrified 10 (8.8%); PHEV only 5; HEV only 1; hydrogen FCEV only 3.
• By vehicle volume: ICE-only 7,800,887 units (57.1%); mixed platforms 5,317,822 (38.9%); BEV-only just 255,502 (1.87%). All-electrified powertrains combined = 4.00% of units.
• Fleet-normalized reversal: BEVs are 1.87% of the ~293.5M-vehicle US fleet but 31% of campaigns — about 6.36 campaigns per million registered BEVs vs 0.215 for ICE (~30x).
• News peg: Lucid 26V540000 (Lucid ref SR-26-06-0), 27,185 model-year 2022-2026 Air; ELECTRICAL SYSTEM:SOFTWARE; exterior-lighting circuit can overheat and increase fire risk; over-the-air software remedy, free; owner letters expected 2026-10-16; park outside until remedied.
• Mechanism caveat: Only 40 of 113 campaigns / 786,979 units (5.8%) involve a fire that originates in a high-voltage/traction battery or hybrid propulsion system. The other 94.2% are conventional systems every gas car shares (ABS/HECU modules, brake fluid, engine, wiring, trailer hitches, seat switches).
• POTAFF is potentially-affected units, an upper bound — not confirmed fires. A single VIN can appear in several campaigns.

Do most 'park outside' fire recalls involve EVs or gas cars?

By raw count, gas cars. Gas/diesel-only campaigns are 52.2% of the 113 campaigns and 57.1% of the 13.67M potentially-affected units, while battery-electric-only campaigns are 31.0% of campaigns but just 1.87% of units. The three largest campaigns in the whole set are mainstream gas platforms: the Hyundai/Kia ABS-module fire recalls (23V651, 1.65M units; 23V652, 1.73M) and the Jeep steering-pump recall (26V363, 1.08M) — these are labeled "mixed" only because a handful of electrified trims share the platform.

That directionally confirms the Washington Post / CleanTechnica framing that park-outside recalls are overwhelmingly an internal-combustion story. It is also incomplete, because it counts vehicles rather than adjusting for how few EVs are on the road. This is a recall analysis, distinct from our earlier car caught fire while parked study, which counted 51,587 consumer complaints of parked fires and never split by powertrain. Our full campaign-by-powertrain breakdown, including every closed campaign back to 2015, also extends the raw enumeration in our how to check if a used car has a park outside fire recall piece, which listed the 68 active campaigns but never split them by powertrain.

If EVs are a tiny share of cars on the road, are they still over-represented?

Yes, at the campaign level — this is the cut that flips the headline. BEVs are about 1.87% of the roughly 293.5M US light-duty vehicles in operation (5.5M of 293.5M, Experian), but they account for 31.0% of these campaigns. Normalize by fleet size and BEVs draw about 6.36 park-outside fire-recall campaigns per million registered vehicles, versus 0.215 for gas/diesel — roughly 30x more often per car on the road. On that measure, Lucid's recall is not a fluke; it is the pattern.

The reversal does not hold on every metric. Measured as units placed under a park-outside recall as a share of each powertrain's own fleet, BEV is about 4.65% versus 2.84% for pure gas (4.78% if you fold in the mixed platforms) — roughly comparable, and the BEV figure is inflated by one saga (see below). So the honest answer is: EVs are far over-represented in how often these campaigns are issued, but not dramatically over-represented in the share of their fleet swept up, and the absolute vehicle counts are still dominated by gas cars.

Is the Lucid Air fire risk coming from the battery?

No. Lucid campaign 26V540000 covers 27,185 model-year 2022-2026 Air sedans, and NHTSA classifies the defect under ELECTRICAL SYSTEM:SOFTWARE: an exterior-lighting circuit has insufficient overcurrent protection and can overheat, and "an overheated circuit increases the risk of a fire." The fault also can disable the center brake light and front lights, reducing visibility. The remedy is a free over-the-air software update; owners are told to park outside, away from structures, until it is applied, with owner letters expected 2026-10-16.

That nuance matters for the whole debate. Even this EV peg is not a traction-battery fire — the source is a headlight circuit, the kind of conventional electrical failure any car can have. It is why we classify 26V540000 as BEV powertrain but conventional mechanism. Across the full set, only 40 of 113 campaigns and 786,979 of 13.67M units (5.8%) involve a fire that starts in the high-voltage battery or hybrid propulsion system; the other 94.2% are ABS/HECU modules, brake fluid, engine, wiring, trailer hitches and seat switches — hardware a gas car carries too.

Which had the biggest EV fire recall — was it the Chevy Bolt?

Yes, and it dominates the EV column. The Chevrolet Bolt EV/EUV LG-battery saga — campaigns 20V701, 21V560, 21V650, 24V481 and 24V812 — accounts for 160,939 of the 255,502 BEV-only units, or 63%. Strip the Bolt out and battery-electric park-outside recalls fall to 94,563 units across the entire 2015-2026 window, which is why BEV's "share of its own fleet recalled" figure should be read with that concentration in mind. This is the one cut that reverses the finding: excluding the Bolt, BEV-only units fall to 1.72% of the BEV fleet, below the 2.84% pure-gas figure. The unit-level comparison is therefore carried by a single nameplate; the campaign-rate comparison is not (see Robustness checks). The Bolt is also a genuine traction-battery case, unlike Lucid's.

The rest of the BEV and electrified column is a long tail of much smaller campaigns — Jaguar I-Pace, Mercedes-Benz EQB, Volkswagen ID.4, BMW i4/iX, Nissan Leaf, Kia EV6/EV9 and Hyundai Ioniq 5 among them. The Ioniq 5 / EV6 / EV9 battery action is covered in our used electric car battery recall check before buying explainer, and one BMW park-outside case is walked through in how to check a used bmw for open recalls before buying.

Which gas cars have park-outside fire recalls right now?

The largest by far are mainstream gas platforms. The Hyundai/Kia ABS-module ("HECU") recalls (23V651 and 23V652) alone total roughly 3.38M vehicles, and the Jeep Wrangler/Gladiator steering-pump recall (26V363) adds about 1.08M — all carrying the same park-outside directive owners associate with EVs. We break down that case in should i buy a used jeep wrangler with an open recall. The rest of the gas column is a long tail of smaller fire-risk campaigns in conventional systems — brake fluid, engine, wiring, trailer hitches and seat switches. The takeaway for a used-car shopper: "park outside" is a gas-car warning at least as often as an EV one, and the biggest single campaigns are gas.

By the numbers

• 113 park-outside fire-risk recall campaigns, 2015-2026 (100 by NHTSA's official flag alone; 109 by keyword; 113 flag-or-keyword).
• 13,666,010 potentially-affected units — an upper bound, not confirmed fires.
• 52.2% of campaigns and 57.1% of units are gas/diesel-only.
• 31.0% of campaigns but 1.87% of units are battery-electric-only.
• ~30x — BEV campaigns per registered vehicle vs gas (6.36 vs 0.215 per million).
• 5.8% of units involve a fire that starts in the traction battery or hybrid propulsion system.
• 63% of BEV-only units are the Chevy Bolt saga alone.

Based on 113 NHTSA park-outside fire-risk recall campaigns (2015-2026, N=113 of 10,354 vehicle campaigns) from the FLAT_RCL_POST_2010 flat file, retrieved 2026-08-31; fleet denominators from Experian vehicles-in-operation (293.5M total, 5.5M BEV). Powertrain inferred from model-name patterns; mixed ICE+electrified platforms are reported as their own bucket, not force-split. Frozen dataset linked in Sources.

Method

Source file. NHTSA Recalls flat file FLAT_RCL_POST_2010.txt (inside FLAT_RCL_POST_2010.zip), downloaded from static.nhtsa.gov/odi/ffdd/rcl/ on 2026-08-31 (file LastModified 2026-08-31T07:04:05Z), 244,925 rows. One row per make/model/model-year within a campaign, so campaign-level figures require aggregation.

Universe. Vehicle recalls only — campaign number position 3 = V, which excludes equipment (E), tire (T) and child-seat (C) campaigns — with a campaign year of 2015 or later. That base is 10,354 vehicle campaigns.

Filter. From that base we kept campaigns that either carry NHTSA's own PARK_OUTSIDE advisory flag (field 29 = Yes) or contain park-outside / park-away-from-structures language in the combined DESC_DEFECT + CONEQUENCE_DEFECT + CORRECTIVE_ACTION text, together with fire language. That yields N = 113 campaigns and 13,666,010 potentially-affected units. The flag alone returns 100 campaigns and the keyword route alone 109; we report the union and disclose all three counts rather than picking the most flattering.

Powertrain classification. Each model string was classified BEV / PHEV / HEV / FCEV / ICE by model-name pattern, then rolled up to the campaign. A campaign whose trims span both combustion and electrified powertrains is labelled MIXED and reported as its own bucket — it is never force-split, because POTAFF is published per campaign and cannot be apportioned across trims. Every MIXED and BEV campaign was then re-checked model-by-model by hand, which corrected five campaigns initially mis-tagged as containing a BEV (notably 15V701000, the 1997-2004 GM sedans, and 25V636000, the BMW 2/3/4/5-series — model names like 330I and IMPALA falsely matched BMW i-series patterns).

Dedup. Campaign numbers are deduplicated to the nine-digit CAMPNO. A single VIN can legitimately appear in several campaigns, and repeat campaigns against the same nameplate are counted separately in the headline figure — see Robustness checks, where we also report the nameplate-deduplicated version.

Normalization. Fleet denominators are Experian vehicles-in-operation: 293.5M US light-duty vehicles, of which 5.5M are BEV (1.87%). Rates are campaigns per million registered vehicles of that powertrain.

Known limits. (1) POTAFF is potentially affected units — an upper bound, not confirmed fires or even confirmed defects. (2) Recall counts measure manufacturer and regulator behaviour as much as underlying risk; a young fleet under heavy regulatory attention will generate campaigns faster than an old one. (3) Model-name classification is imperfect for badge-engineered and fleet models. (4) The numerator contains some medium/heavy and commercial vehicles while the denominator is light-duty — quantified in Robustness checks. (5) 2026 is a partial year.

Robustness checks

We ran the checks most likely to reverse the headline. Three did not; one did, and it is disclosed above.

Nameplate dedup. The 35 BEV-only campaigns collapse to 18 distinct nameplates: the Jaguar I-Pace alone accounts for 7 campaigns and the Chevrolet Bolt 5, so a handful of low-volume models are re-recalled repeatedly. ICE campaigns are less concentrated (59 campaigns, 43 nameplates). Deduplicating cuts the gap from ~30x to 21.9x — it does not close it.

Commercial vehicles. 4 of 35 BEV campaigns and 14 of 59 ICE campaigns are transit buses, Class-8 trucks, LSVs or motorhomes, which do not belong in a light-duty denominator. Removing them from both sides widens the gap to 36.1x, because commercial vehicles are a larger share of the ICE column.

The one reversal. At the unit level, removing the Chevrolet Bolt saga flips BEV from slightly above gas to below it (1.72% vs 2.84%). We therefore do not claim EVs are recalled in greater proportion of their fleet. The claim that survives every cut is narrower and specific: park-outside fire-risk campaigns are issued far more often per registered EV than per registered gas car, while the vehicles swept up by them remain overwhelmingly gas.

What a VIN check can and can't tell you here

A VIN check is a screening step, not a clearance. Zilocar surfaces whether a specific VIN is named in an open recall and the recall count — the same presence-and-count NHTSA's free VIN tool provides — which is useful when 113 campaigns and 13.7M vehicles across every powertrain carry a park-outside warning and you are about to put one in an attached garage. It cannot confirm the fire remedy was performed, which is acute for Lucid's over-the-air fix: "recall present" does not tell you whether that Air already downloaded the patch. Only NHTSA, Lucid, or the dealer can confirm remedy status.

Where Zilocar adds value beyond the free recall lookup is the surrounding history on the same VIN: prior accident and airbag-deployment records, salvage and junk-auction records, odometer/rollback checks, ownership history, and past and current sales-listing history (prices, mileage, days-on-market). Screen for the recall's presence first, then vet the car's real-world history. A Zilocar VIN check bundles the recall-presence screen with that accident, salvage-auction, odometer, ownership and listing history in one report; it does not confirm a remedy, track investigations, or show firmware status — leave those to NHTSA and the dealer.

Related checks

• How to check if a used car has a park outside fire recall
• Used electric car battery recall check before buying
• Should I buy a used Jeep Wrangler with an open recall

Sources

• NHTSA Recalls flat file — FLAT_RCL_POST_2010.zip (primary dataset; the retired FLAT_RCL.zip path is dead)
• NHTSA Recalls flat file record layout — RCL.txt (field definitions, incl. PARK_OUTSIDE advisory flag)
• NHTSA Part 573 defect report — Lucid 26V540000 (RCLRPT-26V540-2255.pdf)
• NHTSA Recalls API — recallsByVehicle (Lucid Air)
• NHTSA datasets & APIs
• Experian Automotive Market Trends Q3 2025 — vehicles in operation (293.5M total; 5.5M BEV, 11.7M hybrid)
• U.S. EIA — EV sales fell as hybrid sales rose in 2025
• CNBC — Lucid to recall more than 27,000 luxury sedans over fire risk
• InsideEVs — Lucid recalls 27,185 Air sedans over fire risk
• Frozen campaign-level dataset (this analysis, powertrain split)

Last verified: 2026-08-31