The register tracks a few thousand component names, and most of them lead quiet lives. Brake hoses fail, get recalled, get replaced, and nobody writes books about them. One family does not have a quiet life. The air bag entered American cars as the part that would save you, and then proceeded to write the loudest file in the building: roughly 151,000 complaints, 1,064 distinct recall campaigns covering just under 150 million units, a word that had to be added to the register's working vocabulary, the largest recall event ever run, and, in one decade, a third of every death reported to the complaint register. I pulled the family's complete file from both registers, the recalls and the complaints, plus the investigation table. This is the biography, told entirely from the paperwork.
Page one: February 1990
The complaint register has a prehistory for this family. A thin scatter of air-bag complaints carries failure dates from the 1970s and 1980s, the era when a few experimental cushions rode in big sedans, and an even thinner scatter claims dates from the 1960s, before the part existed in any showroom, which tells you those entries are clerical back-filing rather than history. The recall file has no prehistory at all. The 1970s and 1980s hold 5,510 campaigns between them and not one of them is filed under the family. Then the mandate era arrives, air bags start shipping in volume, and the file opens properly on February 12, 1990: campaign 90V031000, Buick Reatta, a forward sensor that may have been built with part of the wiring harness in the wrong place.
The second page is the one worth framing. April 20, 1990, Toyota Celica: "air bag inflator case was incorrectly machined." Five days later, Chrysler recalls Dodge Dynastys, LeBarons and Daytonas because the air bag inflator modules may not contain diffuser holes between the ignitor and the propellant chamber. The register's second and third entries for this family, filed within a week of each other in the spring of 1990, are both about the metal canister of propellant at the center of the device. The part that would make the family famous twenty-three years later was already the fragile heart of it on page two. The early file stays charming for a while: a 1991 BMW campaign because an incorrectly installed knee bolster was discovered during testing, and a December 1991 Toyota MR2 recall because the driver's protection system exceeded the 60g thorax criteria, a sentence in which the safety equipment itself fails a crash test. The 1990s close with 72 distinct air-bag campaigns covering about 4.8 million units, a modest opening chapter.
The quiet majority
Ask anyone what goes wrong with an air bag and you will hear about the famous failure, the one with the shrapnel. The file disagrees. I ran every air-bag complaint narrative through four symptom buckets, and the most common thing an owner reports about this family, by a clear margin, is a light on the dashboard. About 19.4% of all air-bag complaints ever filed describe a warning light. Another 16.1% report the opposite of the famous failure: a bag that did not deploy in a crash. Unwanted deployment, the bag going off on an ordinary Tuesday, is 1.8%. And rupture or shrapnel, the failure that made the family a household name, appears in 1.7% of the family's complaints. The most dramatic failure in the register is also the rarest of its four voices.
The decades sharpen this. Rupture language barely exists in the early file, 63 narratives in the 1990s and 214 in the 2000s, then jumps to 1,719 in the 2010s as the crisis surfaces. But the same decade produced 17,760 warning-light narratives, ten for every shrapnel report. The Takata era is visible in the complaint register less as explosions than as a nation of owners staring at an amber light and filing a federal complaint about what it might mean. Fear of the failure generated far more paperwork than the failure.
The harm columns
Every complaint in the register carries two fields most articles on this site have never needed: the number of people the complainant says were injured, and the number they say died. For most component families those columns are almost all zeroes. For this family they are not. Air-bag complaints carry about 39,500 reported injuries and more than 1,000 reported deaths across the file. The decade cut is the finding. In the 1990s, air-bag complaints carried 12.3% of all deaths reported to the register. In the 2000s, 11.9%. In the 2010s, air-bag complaints carried 500 of the 1,550 deaths reported that decade: 32.3%. One drawer of the parts bin, a third of the reported deaths. These are complainant-reported figures, not adjudicated ones, and the register makes no promise that every number in those columns survived an investigation. But the register also does not need the numbers to be perfect to show the shape: no other component family owns a decade of the harm columns the way this one owns the 2010s.
Takata, counted properly
The recall table stores one row per make and model, which means the crisis looks even bigger in the raw file than it was: thousands of rows. Collapse the rows into distinct campaign numbers and the honest count emerges, and it is still enormous. There are 133 distinct campaigns in the air-bag drawer whose filings carry the rupture signature, Takata by name or an inflator that can rupture. Deduplicated to one unit count per campaign, they cover 48.3 million units. For scale, the family's entire file, 1990 to now, covers about 150 million units across 1,064 campaigns, so this one defect accounts for roughly a third of every air-bag unit ever recalled. It is also the engine of the family's worst decade by volume: the 2010s alone hold 544 air-bag campaigns covering 105 million units, seven of every ten units in the family's history, recalled in a single decade. The 2010s are to this family what the largest-recalls ledger is to the register as a whole.
Two things about the campaign list surprised me. First, the earliest filing in the drawer that carries the rupture signature is not from the crisis years at all. It is a 2003 General Motors campaign, 03V051000, covering Buicks, Cadillacs, Chevrolets, GMCs and Pontiacs for a driver-side inflator, a full decade before the named campaigns begin arriving in 2013. The register contained the signature before it contained the scandal. Second, the breadth. The named campaigns peak in 2016 and 2017, 68 distinct campaigns in two years, and they spread across dozens of makes with no single make holding more than 17 of the 133. This was never one company's problem, because the economics of the part made one supplier's canister everyone's canister. And the chapter has no last page: new campaigns carrying the signature were still being filed in 2024, 2025 and 2026. The crisis is old enough to drive and the paperwork is still arriving.
A word enters the register
You can watch the vocabulary change. The word "inflator" appears in 0.12% of 1990s recall filings and 0.32% of 2000s filings, a technician's word for an internal part no owner ever thought about. In the 2010s it appears in 5.97% of all recall filings in the register, every component family included. One part of one device became, for a decade, a measurable share of everything the regulator wrote. The complaint register never followed: owners file symptoms, lights and bangs and bags that stayed shut, and it is the makers and the regulator who file the part. The same taxonomy lag runs the software story in the Parts Bin, and the inflator is its sharpest single-part example.
Six dockets, a hundred and forty-five thousand rows
The investigation table looks, at first glance, like the family's biggest file of all: 145,336 rows, the overwhelming bulk of them air bags in the 2010s. Collapse to distinct investigation numbers and the truth is almost the opposite. The regulator has opened 179 distinct air-bag investigations in the file's entire history: 54 in the 1990s, 64 in the 2000s, 51 in the 2010s, 9 so far in the 2020s. Six of the 179 mention Takata. Those six dockets account for roughly 96% of every air-bag row in the table, because a docket that touches hundreds of models across dozens of makes explodes into tens of thousands of vehicle rows. Six case numbers produced more paperwork than every other air-bag investigation in history combined, which is the tidiest one-sentence portrait of the crisis the file offers.
The twenty worst files
Rank every vehicle in the register by air-bag complaints alone and the top of the table is its own little museum. Three Ford Fusions hold three of the top four slots. Six different Toyota Corolla model years, 2005 through 2011, make the twenty, a wall of the exact cars that sat at the center of the inflator era. A 2002 Jeep Liberty sits in the top six. Every count below clears the site's percentile ladder at the 96th percentile or higher on air-bag complaints alone, before a single other component is added.
The family goes quiet
The last chapter, so far, is a return to something like a normal life. The family's share of the complaint register, which ran a touch over 5% through the 1990s and 2000s and spiked to 11.1% in the 2010s, has fallen to 4.0% in the 2020s. Its share of reported deaths is down to 8.2%. The recall side has not gone quiet in the same way, 307 distinct campaigns and nearly 30 million units already this decade, but a growing share of that is the long tail of the old defect being chased through old cars rather than a new one being discovered.
So the biography, in one paragraph. Born into the register in February 1990 with a miswired sensor on a Buick. An inflator defect by page two. A modest first decade, a famous middle age in which one supplier's canister turned the family into a third of the register's reported deaths and six dockets into a hundred thousand pages, and a present in which the loudest thing the family does, most days, is turn on a light that 29,360 people have formally told the government about. The part that was installed to end the worst column in the file instead spent one decade owning it, and the file, being a file, simply wrote that down too. The newest campaigns carrying the old signature arrived this year. Biographies of parts do not end, they just go quiet, and this one is quieter than it has been in twenty years.
Notes on the cut. Air-bag rows were selected by component name in both registers. Harm figures come from the complaint register's injured and deaths fields, which are complainant-reported and not adjudicated. The Takata set was identified by filing text, the supplier's name or an inflator-rupture description, so a campaign described without either would be missed. Recall unit counts take one potentially-affected figure per distinct campaign number. Complaint failure dates before 1990 are sparse and partly back-filed, and are read as texture, not as trend.