GM Studied 12 Million of Its Own Cars. It Won't Tell You What It Found.

In May 2026, General Motors published its eighth joint safety study with the University of Michigan Transportation Research Institute. Dataset: 12 million MY2020-2024 GM vehicles matched to 700,000 police-reported crashes across 18 states. GM's conclusion: their safety technologies deliver "statistically significant" reductions in crashes and injuries.^[1] The actual crash reduction percentages? Not in the release. Not in any public document GM has made available.

Read the press release carefully. Twelve paragraphs of quotes from GM executives. Two mentions of "striking." Three of "statistically significant." The phrase "high degree of confidence" appears once. An infographic showing the results was published as an image, not as downloadable data. Not once does the word "percent" appear in the text. Eight studies over multiple years, funded by GM, executed by a respected university, and the public gets adjectives.

Compare that to how the Insurance Institute for Highway Safety publishes its research. When IIHS found that electronic stability control reduced single-vehicle fatal crash risk, it said 56%. When it measured the effect of AEB on rear-end crashes, it published the number: 50%. These figures are cited in federal rulemaking, insurance pricing, and consumer decisions because they are numbers, not vibes.^[2]

GM's study covers AEB, front pedestrian braking, lane keep assist, forward collision alert, and IntelliBeam. These are standard on GM vehicles under $30,000 MSRP, including the Trax, Trailblazer, and Bolt. The company says every one of them works. It will not say by how much.

So we ran our own comparison using FARS. GM vehicles from model years before 2020 accumulated 45,853 fatalities across 872 model-year entries in the FARS 2014-2023 dataset. Model years 2020 and later: 1,055 fatalities across 56 entries. Average per entry: 52.6 pre-2020, 18.8 post-2020. A 64% reduction. Then we ran Toyota. Pre-2020: 54.0 average, post-2020: 26.3. A 51% reduction. Honda: 99.0 to 22.6, a 77% reduction.^[3]

Every major manufacturer's newer vehicles kill fewer people per model-year cohort. Honda's reduction is larger than GM's. Toyota's is smaller. None of these comparisons prove that any specific technology caused the improvement, because newer vehicles have fewer cumulative miles on the road during the FARS window. A 2023 model had one year of exposure; a 2015 model had eight. Fleet age dominates every comparison that does not control for mileage, and FARS does not report per-vehicle mileage.

GM knows this. UMTRI certainly knows this. Its study methodology, described only in general terms in the press release, says researchers compared "system-relevant crashes to control crashes" for vehicles with and without specific features, controlling for "driver behavior, road conditions, and vehicle type." That is a reasonable approach, possibly a rigorous one. But without the actual numbers, the methodology is a recipe with no dish.

Meanwhile, the Silverado comes standard with every technology GM credits. It has 9,591 fatalities in FARS and a death rate of 1.25 per 100 million VMT. The Traverse, with the same safety suite, sits at 0.20. Same company, same technologies, a 6.25x gap in death rate. The difference is not AEB versus no AEB. It is vehicle type, mass, usage pattern, and who is driving it.^[3]

If you are evaluating a vehicle's safety: look at IIHS ratings and NHTSA star ratings, both of which publish actual test data. Manufacturer-funded studies that omit their own findings are marketing documents with a university letterhead. GM's safety tech probably works. Make them prove it with numbers.