Congress Ordered NHTSA to Stop Hot Car Deaths in 2021. The Rule Is 2.5 Years Late. Three Children Died Last Week.

Three children died in hot cars on a single day last week.^[1] A 4-year-old in Los Angeles, a 2-month-old in Fredericksburg, Virginia, left by her mother while she worked a shift, and a 1-year-old in Brookwood, Alabama, where the temperature hit 90 degrees and the inside of a parked car can exceed 130 within half an hour. Temperatures across the three cities ranged from 87 to 97 degrees. Five children total in 2026, and summer hasn't started.

This number, compiled by San Jose State University's Department of Meteorology and Climatology and maintained at noheatstroke.org, has climbed by roughly one child every ten days for 28 years.^[2] It peaked at 53 in 2018 and 52 in 2019, dropped during the pandemic when more parents worked from home and fewer toddlers were shuttled to daycare on autopilot, and is now rebounding as routines snap back to normal. Thirty-one died in 2025.

In November 2021, Congress passed the Infrastructure Investment and Jobs Act, and Section 24220 directed the Department of Transportation to issue a final rule requiring rear-seat occupant alert systems in all new passenger vehicles.^[3] The congressional deadline for that final rule was November 2023. NHTSA blew past it without publishing even a proposed rulemaking, then estimated it would issue a proposal by April 2025.^[4] That deadline also passed. Nothing exists. No proposed rule. No comment period. No timeline. As of this writing, a law signed by the President four years ago has produced zero pages of regulatory text.

The math I want to show you requires exactly two inputs and one assumption, and it will make you angry. Input one: 37 children die in hot cars per year on average. Input two: the rule is 2.5 years overdue. The assumption, generous to NHTSA, is that the technology only works in new vehicles, fleet penetration starts at zero, and the system prevents 80% of deaths in equipped cars. If the rule had been finalized on time, model-year-2025 vehicles would carry in-cabin occupant detection, roughly 17 million new cars sold per year would mean about 34 million equipped vehicles on the road by now, covering 11.7% of the 290-million-vehicle U.S. fleet, and at 37 deaths per year with that penetration and effectiveness rate, the system saves 3.5 children per year by 2026 alone. Compound it. By 2030, penetration reaches 41%, and the system prevents 12 deaths annually. Through 2035, the cumulative excess death toll from this single regulatory delay reaches an estimated 50 to 80 children who would have been saved had the rule shipped on schedule.

Fifty to 80 is a range, not a count, and I owe you the caveats before anyone quotes the upper bound at a congressional hearing. Fleet turnover is estimated from IHS Markit registration data, and "80% effectiveness" is a projection based on Vayyar's claimed detection rates, not a field-validated number.^[5] The real effectiveness could be lower. It could also be higher: radar-based detection doesn't rely on the driver remembering anything at all but instead measures the presence of a breathing body in the cabin and triggers the alarm whether or not the driver follows a routine. The floor matters more than the ceiling here, though. Even at 60% effectiveness and slower fleet turnover, the delay costs at least 35 children through 2035.

The technology is not hypothetical: in 2021, the FCC granted waivers for in-cabin radar operating in the 60 GHz band to Infineon, Tesla, Valeo, and Vayyar.^[6] Euro NCAP has required child presence detection for its five-star rating since 2023. Twenty-plus automakers, including GM, Hyundai, and Toyota, voluntarily committed to rear-seat reminder systems by model year 2025.^[7] Some delivered: many 2025 models chime when the rear door was opened before a trip but not after. That door-sequence logic is better than silence, but it cannot detect a child who climbed into an unlocked car on their own, which accounts for roughly 25% of hot car deaths.

Radar can, and it costs about $50 per vehicle. On 17 million new vehicles per year, that is $850 million annually spread across an industry that generated $1.1 trillion in U.S. sales in 2024. NHTSA's own enabling statute, signed into law by a bipartisan Congress, told the agency to mandate it. Two and a half years later, the agency hasn't managed to write the first draft.

Do not confuse this with the separate rear seatbelt warning rule NHTSA finalized in January 2025, which requires audible alerts when rear passengers are unbuckled.^[8] That rule addresses seatbelt use, not occupant detection, and compliance isn't required until September 2027. The two rules get conflated in press coverage, but they solve different problems.

I ran the delay-body-count model at three effectiveness levels (60%, 80%, 95%) and two fleet turnover speeds (baseline 17M/year and accelerated 19M/year reflecting EV transition). All six scenarios produce the same structural conclusion: the relationship between regulatory delay and cumulative excess deaths is superlinear because each year of delay shifts the entire fleet penetration curve to the right, and the area under that shifted curve grows with time. A one-year delay doesn't cost one year of deaths. It costs an expanding wedge of deaths that persists for the entire fleet replacement cycle.

The strongest case against this analysis: the voluntary automaker commitments may already be reducing deaths without a federal mandate. At 31 deaths, 2025 came in below the 37-per-year average. Maybe door-sequence alerts are working. But 31 is still 31, and the 2026 pace (5 dead by late May) projects to 24 for the full year only if the summer months, when the vast majority of hot car deaths occur, hold to the pre-summer rate. They won't. June through August accounts for roughly 75% of annual hot car fatalities. If 2026 follows historical seasonal patterns, the final count will be closer to 35.

On May 21, the temperature in Fredericksburg was 97 degrees. Inside a parked car at that ambient temperature, the cabin reaches 130 degrees within 30 minutes and 150 degrees within an hour. A 2-month-old infant has a surface-area-to-mass ratio roughly three times that of an adult. Core body temperature rises faster, and organ failure begins sooner. The window between "left in the car" and "dead in the car" can be as short as 15 minutes.

A $50 radar sensor, pulsing 60 GHz through the cabin once per second, would have detected that infant's breathing. It would have triggered the horn, flashed the lights, sent a push notification. NHTSA was told to mandate this 2.5 years ago. The proposed rulemaking doesn't exist.