Why portable generators kill during outages

I work on battery systems, so I have an obvious interest in this comparison, and I am going to state my position at the top rather than let it leak out of the argument: I help build AI for a company that sells energy storage. Read the numbers below as the government's, not mine. Every figure here comes from two federal documents that anyone can open.

What made me want to write this is not that generators are dangerous, which everyone knows in a vague way. It is that the danger is quantified, published annually, and almost entirely absent from the buying conversation. People compare a generator and a battery on watts and dollars. The strongest differentiator is a toxicology fact, and it never appears on the spec sheet.

The number that explains the rest

In its 2023 proposed safety standard, the Consumer Product Safety Commission writes this:

The CO emission rates of portable generators are on the order of hundreds of times the CO emission rates of gasoline powered automobiles.

That one sentence reframes every piece of generator safety advice you have ever read.

Most people's intuition about engine exhaust is calibrated on cars. A car idling in an open driveway feels survivable, because it is. If you carry that intuition to a generator, you will make a placement decision that is off by two or three orders of magnitude in source strength. Small engines built to a price point, without the catalytic aftertreatment and closed-loop fuel control that cars have carried for decades, are simply a different kind of emitter.

This is why "run it outside" is not a sufficient instruction and why so many deaths happen with the generator technically outdoors, in a detached structure, under an overhang, or in a garage with the door open. Ventilation that would comfortably dilute a car does not comfortably dilute this. The gas is colorless and odorless, so the feedback loop that normally protects people, noticing and leaving, is missing. CPSC notes that with rapidly rising CO levels of the kind produced by gasoline engine-driven tools, sudden extreme hypoxia can cause rapid incapacitation and loss of consciousness, which prevents an exposed person from leaving.

You do not get a warning and then a decision. You get a decision you are no longer able to make.

What the fatality data actually says

The CPSC's report Non-Fire Carbon Monoxide Deaths Associated with the Use of Consumer Products, 2022 Annual Estimates was published in May 2026. For 2022 it estimates 274 unintentional, non-fire CO poisoning deaths from consumer products under the agency's jurisdiction, drawn from 182 reported incidents.

The breakdown is the interesting part:

• Engine-driven tools were the largest category, associated with an estimated 107 deaths, 39% of the total. Counting incidents that involved multiple products, engine-driven tools reach an estimated 129 deaths, 47%.
• Of those 107, an estimated 93 involved generators. CPSC calls generators "the single product under CPSC's jurisdiction that is associated with the most CO deaths."
• Heating systems were second, at an estimated 76 deaths, 28%.
• Since 2012, generators alone are associated with an estimated 892 non-fire CO poisoning deaths, 41% of all consumer-product CO deaths in the agency's scope.
• Staff found a statistically significant upward trend across 2012 to 2022, and the estimate peaked in 2022, the final year of the series.

Two further patterns matter for anyone planning backup power.

It is a home problem. An estimated 174 of the 274 deaths in 2022, 64%, involved exposure in a home location. Within that, an estimated 34 deaths occurred in detached structures at residential locations such as sheds and detached garages, and a further 14 in structures not designed as permanent residences. Across 2020 to 2022 the annual average was 174 deaths in home locations, 74% of the all-cause average of 238.

It is an outage problem. In 2022, an estimated 174 deaths, 64%, occurred in the four coldest months of November through February, against an eleven-year average of 55% for those months. CPSC states the mechanism directly: portable generators are often used in the cold months because of power outages due to snow and ice storms.

That is the shape of the hazard. It concentrates in homes, during outages, in exactly the scenario the generator was bought for. The device is most dangerous at the moment of its intended use, which is an unusual property for a consumer product and the reason it deserves a place in the buying decision.

The rule that has not arrived in twenty years

In April 2023 the CPSC published a supplemental notice of proposed rulemaking, Safety Standard for Portable Generators, at 88 FR 24346. It would limit CO emissions from portable generators and require them to shut off when specified levels are reached. The docket number is CPSC-2006-0057. The rulemaking has been open since 2006.

The Commission's own case for it is strong. From 2004 through 2021 it counts at least 1,332 CO-related consumer deaths involving portable generators, an average of about 74 per year, and for the three most recent years of complete data at that time, 2017 through 2019, generator-related CO deaths averaged 85 per year. Over 30 years the Commission estimates the rule would prevent 2,148 deaths, nearly 72 per year, and 126,377 injuries, with total benefits above $1 billion per year and net benefits around $897 million per year at a 3% discount rate, roughly $273 of benefit per generator sold.

Comments closed on 20 June 2023. I queried the Federal Register for CPSC documents concerning portable generators; the most recent is from June 2023, and there is no final rule. So as of today the standard is voluntary, and a generator on a shelf this weekend is not federally required to shut itself off when it is poisoning someone.

There is a detail inside the proposal that I think is the most useful thing in it for a buyer, and it cuts against a marketing claim. The SNPR adopts emissions limits in addition to shutoff requirements, and explains why:

both actual fatal incidents and scenario simulations show that an effective shutoff system alone is not sufficient to protect consumers from death and serious injury from accumulated CO.

A generator advertising a CO shutoff sensor is meaningfully better than one without. It is not, on the agency's own analysis, sufficient. The sensor responds to accumulated gas; by then a large quantity has been produced. If you are shopping, the relevant voluntary standards named in the proposal are UL 2201 and PGMA G300, and the honest question to a seller is which standard the unit is certified to, not whether it "has a CO sensor."

The comparison I am not going to overstate

Here is where I owe you the counterweight.

A battery system produces no combustion products. That is not a marketing claim, it is chemistry: there is no fuel and no exhaust, so the carbon monoxide pathway does not exist. On this specific hazard the comparison is not close.

It would be dishonest to stop there, because batteries have their own failure mode and I have written about it repeatedly. Lithium systems can fail thermally, and I have covered why "UL 9540A certified" is not a thing anyone can actually be, what the falling battery fire rate does and does not tell you, and the chemistries that trade energy density for not burning. A battery installed badly, in the wrong place, without regard to the installation standards, is a real hazard. The correct summary is not "batteries are safe and generators are not." It is that the two technologies fail in different ways, and only one of them fails by filling your house with an odorless gas during the storm you bought it for.

I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator; I do not own the company and earn nothing from this link. Full policy here. That is exactly why I have kept every number in this report sourced to the CPSC rather than to anyone selling anything, including us.

Cost, runtime and refuelling all still favour a generator in plenty of real situations, particularly for long multi-day outages where a battery runs flat and a fuel can does not. This report is not an argument against owning one. It is an argument for pricing the hazard into the decision instead of leaving it out.

If you already own one

The mitigations are not exotic, and they are worth more than the purchase decision for most readers.

Put CO alarms in the house, on every level and near sleeping areas, and test them before storm season rather than during it. The alarm is the only part of this system that gives you the warning the gas itself does not. Never run a generator inside any enclosed or partly enclosed space, and treat an attached garage with the door open as enclosed, because it is. Keep it well away from doors, windows and vents, downwind if you have a choice, and remember that "outside" is a starting condition rather than a guarantee once wind and building pressure get involved.

CPSC publishes specific placement distances in its consumer guidance. I am deliberately not quoting a number here, because its safety education pages returned 403 to me and I do not restate a safety-critical figure I could not open myself. Get it from the agency directly, and from the label on your unit.

What I could not confirm

The fatality figures are estimates produced by CPSC's statistical methodology from death certificates and incident files, not a direct count, and the agency marks its five most recent years as incomplete and subject to upward revision. The most recent complete-year data available is 2022, published in May 2026. That four-year lag is inherent to the data source, and it means the current annual toll is unknown rather than low.

I did not open the CPSC staff briefing package underlying the benefit estimates, so the 2,148-deaths-prevented figure and its cost analysis are reported as the Commission stated them in the Federal Register rather than independently checked. I did not read the comments filed on the 2023 proposal, and I have no information about why it has not been finalised.

I did not verify manufacturer-level compliance with UL 2201 or PGMA G300, or how many units on the market today carry either certification. I looked; the sources I found were retailer copy rather than testing data, so I left it out.

No claim here rests on my own research. My work is in microwave spectroscopy and battery systems, not toxicology or epidemiology.

The signal

Backup power is bought under a specific emotional condition: the lights just went out, or a forecast says they are about to. That is a bad moment for a careful risk comparison, which is why the comparison should happen now, in the dry.

The federal government has spent twenty years trying to require that these machines turn themselves off before they kill someone, and has not finished. In the meantime the protection available to you is not regulatory. It is a CO alarm, a placement decision made with the right mental model of the source, and an honest accounting of what each backup option actually risks. The generator's failure mode is not that it runs out of gas. It is that it works perfectly, in the wrong place, while everyone inside is asleep.