IIHS Crowned a Kia and a Tesla This Week. BMW's Icon Failed, and the Morgue Agrees

Here is a fun fact that will ruin your morning commute. The two safest new vehicles in America are a Kia and a Tesla, according to the people who smash cars into walls for a living. BMW's 3 Series, the sedan that invented the yuppie, just failed their test. Ten years of federal death records say the 3 Series kills at 91 times the rate of the Tesla Model Y^[1][4].

IIHS published its latest 2026 ratings wave on September 3. Seven vehicles entered the lab, and two left with Top Safety Pick+, the institute's highest award: the redesigned 2027 Kia Telluride and the 2026–27 Tesla Model Y.^[1] The other five, the BMW 3 Series and i4, the Lexus IS, the Toyota Crown Signia, and the GMC Canyon, qualified for nothing.^[1] Five prestige badges walked in. Nobody's country club parking lot saw this coming.

We ran every nameplate in that wave through a decade of NHTSA fatality data, and the morgue sorted them into nearly the same order as the lab. Model Y: 0.03 deaths per 100 million miles. Telluride: 0.04. Canyon: 0.59, about twenty times the Model Y, and Lexus IS: 0.87, twenty-nine times. BMW 3 Series: 2.73, with 1,237 dead, four times the median sedan.^[4] Two independent institutions, one smashing cars and one counting bodies, converged on the same hierarchy, which almost never happens, so enjoy it.

Pause for the Canyon, the only Detroit Three entry in the wave. It scored "marginal" in the updated moderate-overlap test and "poor" in the vehicle-to-vehicle crash prevention test, which measures what your truck does to everyone else.^[2] GM spokesman Mark Lubin replied that the company is "confident in the safety" of the Canyon and cited its four-star NHTSA rating.^[2] Both claims can be true at once. One rating asks whether you survive your crash. The other asks whether the car next to you survives yours, and you can guess which one the commercials mention.

Now the wrinkle that makes the universe laugh. The Telluride's award applies only to units built after June 2026, because Kia quietly upgraded the headlights mid-year.^[1] Buy the wrong build month and your award-winning SUV is technically a different car. Somewhere a used-car salesman is explaining this to a customer, and neither of them knows it.

An honest objection, stated at full strength: this is not a controlled experiment. IIHS tested 2026 and 2027 redesigns, while the fatality data covers 2014 through 2023, so the morgue judged the previous generations, not the cars in the lab. Model Y and Telluride rates rest on 57 and 31 deaths, thin counts on young fleets that flatter every number. And the 3 Series figure is soaked in driver behavior: 22.1 percent of its dead drivers were impaired, versus 17.6 percent for the Model Y.^[4] Some of that 91x is the car. Some of it is who buys it and what they drink first. No dataset on Earth separates the two.

More limits, because the void demands honesty. FARS counts only fatal crashes on public roads within 30 days, so injuries, parking lots, and slow ICU deaths are invisible. Mileage denominators are estimates, not odometer readings, which makes low-volume rates wobble. The BMW i4 and Crown Signia have no FARS history at all, so their failures stand on lab data alone.^[4]