Your Garage Was Built to Contain a Gasoline Fire. You Are Charging a Lithium-Ion Battery That Burns at Twice the Temperature and Reaches Flashover in Twelve Seconds.
Open IRC Table R302.6. Half an inch of gypsum board on the garage side of the framing, a 20-minute fire-rated door with a self-closing device, sheet-steel protection on duct penetrations. That is what separates your attached garage from the room where your children sleep, and it is not a fire-resistance-rated assembly, because no fire test is required, because the code was written decades ago for one specific scenario: a gasoline vehicle catches fire, the gasoline burns at roughly 600°C, the fire develops over minutes, and half an inch of drywall buys your family time to evacuate.
That scenario no longer describes what is in your garage. Not even close.
FSRI and FDNY researchers sealed e-scooters into residential-scale rooms and triggered thermal runaway in the lithium-ion batteries. What they recorded, published in Fire Technology in 2025, should change how you think about that wall. First visible smoke to battery explosion and window failure: twenty seconds. Flashover within thirty seconds. Room of origin immediately fatal. The temperatures exceeded 900°C, which is roughly 50 percent higher than gasoline combustion, and the gases included hydrogen fluoride, a compound that IRC R302.6 does not mention, that your smoke detector does not measure, and that attacks human lung tissue at concentrations your nose cannot smell.
Fifteen to twenty minutes is how long half an inch of regular gypsum board resists a standard fire curve. Against a lithium-ion thermal runaway event that reaches flashover in 12 to 30 seconds, you are not comparing minutes to minutes. You are comparing a wall designed for a slow burn against something closer to a small explosion, and the wall was never tested against it, because the code never imagined a homeowner would store this chemistry on the other side of it, because in 2000 they wouldn't have, because the e-bike did not exist yet.
What FSRI Proved About Sprinklers, and Why Most Jurisdictions Ignore It
FSRI ran the same experiment with NFPA 13D residential sprinklers installed. Published in a companion Fire Technology paper, the results were unambiguous. Sprinklers activated within 3 to 5 seconds of thermal runaway gas cloud ignition, prevented flashover in every single experiment, and contained the fire to the e-scooter seating area. Charley Fleischmann of FSRI: without sprinklers, the fire places every occupant in danger within seconds and threatens the building soon after.
Sprinklers work, which makes what happened next difficult to explain.
IRC Section R313 has required residential fire sprinklers in new construction since the 2009 code cycle, and most U.S. jurisdictions have since repealed or amended the requirement through local exemptions, responding to builder lobbying on cost grounds, which is understandable until you look at the numbers. NFPA estimates residential sprinkler installation at $1.35 per sprinklered square foot in new construction, which for a 2,400-square-foot home comes to $3,240, less than a kitchen appliance upgrade, less than the annual insurance premium increase you will pay after a garage fire claim. Phoenix Fire Department alone investigated 71 lithium-ion battery fires through November 2025. Garages and sheds accounted for 21 of them. Over half occurred when the battery was not even charging.
The Detection That Exists and the Code That Doesn't Require It
Researchers at multiple institutions have demonstrated machine-learning systems capable of detecting thermal runaway not minutes but hundreds of seconds before flames appear. An SSRN preprint from January 2026 described BTRNet, a deep sequence fusion model combining 1D-CNN and Transformer encoder architectures, which predicted thermal runaway with a temperature prediction error below 0.106°C and issued three tiers of warnings, from minor anomaly to imminent runaway, with enough lead time to shut down a battery, activate suppression, or evacuate a room. A separate team published in World Electric Vehicle Journal an XGBoost framework targeting 5-to-20-second prediction windows, computationally cheap enough for embedded systems.
Every one of these systems was designed for electric vehicle battery management, and not one targets the residential garage, the room where Americans actually store and charge their e-bikes, power tools, and portable power stations.
CPSC's own data tells the incidence story. Between 2019 and 2023: 227 e-mobility battery incidents, 39 deaths, 181 injuries, and Commissioner Richard Trumka acknowledged that number is a vast undercount based on New York City data alone. NYC tracked 268 lithium-ion fires in 2023 and 277 in 2024. Trumka proposed a mandatory UL 2849 safety standard for e-bike battery systems in April 2025. It addresses battery quality and says nothing about the room.
What You Can Do With a Code That Has Not Caught Up
If you are building or renovating a home with an attached garage, IRC R302.6 is a floor, and you should treat it that way. Specify NFPA 13D sprinklers in the garage even if your jurisdiction does not require them, because the research proving they prevent lithium-ion flashover is already published and your insurer will notice before your building department does. For e-bikes and e-scooters, a detached outdoor charging station eliminates the question entirely: no lithium-ion inventory inside the attached structure, no fire separation problem to solve. Never charge lithium-ion devices unattended overnight in an enclosed space connected to your living area.
Battery monitoring devices with thermal anomaly detection exist commercially, though none carry residential building code listing. The technology to detect thermal runaway 5 to 300 seconds before flame onset is peer-reviewed, demonstrated, and sitting in journals that code committees have not yet read. What is missing is the regulatory bridge from laboratory validation to a line item in Table R302.6, and until someone builds it, your garage wall is protecting you from a fire it was never designed to face.
The Counterargument Worth Taking Seriously
Most lithium-ion fires involve uncertified, physically damaged, or counterfeit batteries rather than UL-listed products used as intended. NYC's enforcement of UL 2849 certification for e-bike batteries correlated with a 67 percent reduction in fire deaths from 2023 to 2024, from 18 deaths to 6, which is dramatic and real and suggests that battery quality regulation may be the more effective lever than building code revision. CPSC's proposed mandatory standard targets the battery itself. If battery certification works at scale, the fire code gap shrinks as the hazard does, and mandating upgraded garage fire separation would add several thousand dollars in construction cost against a problem that better battery standards might eventually solve.
Fair point. But the FSRI experiments did not use counterfeit batteries, only standard, commercially available, certified products, and they still reached flashover in seconds, because thermal runaway from manufacturing defect, charging fault, or physical damage does not check the UL listing before it ignites. Battery certification reduces how often it happens. It cannot change the physics of what happens when it does, and the physics says 900°C on the other side of half an inch of drywall that was rated for 600.
Limitations
This analysis relies on FSRI experiments conducted with specific e-scooter battery configurations in controlled research settings. Real garage environments vary in volume, ventilation, stored materials, and battery chemistry across manufacturers. The 15-to-20-minute fire resistance estimate for half-inch gypsum board is approximate and assumes standard exposure conditions; actual performance under lithium-ion thermal runaway, which produces both extreme temperatures and pressure, has not been independently tested against the specific assembly described in R302.6. The AI detection systems reviewed were validated on laboratory datasets, and their real-world performance in residential environments with variable temperature baselines, humidity, and sensor degradation over years of installation remains uncharacterized. Cost estimates for residential sprinkler installation are NFPA averages and will vary by region, contractor availability, and local plumbing codes.