Insurer Swiss Re and Waymo find 88% fewer property-damage claims for robotaxis

December 2024 — Waymo has a new way to argue its cars are safer than people: insurance claims. Here's what the study found, what it cost, and what went wrong.

Cost and outcome (as of Dec. 19, 2024)

Miles studied: 25.3 million fully autonomous miles in Phoenix, San Francisco, Los Angeles and Austin. (Waymo)
Against all human drivers: 88% fewer property-damage claims and 92% fewer bodily-injury claims. (Waymo)
In raw numbers: 9 property-damage and 2 bodily-injury claims, against an expected 78 and 26 for human drivers over the same distance. (Waymo)
Against newer cars ("latest-generation" models with driver-assist): 86% and 90% fewer. The paper gives the model years as 2018–2021 in its abstract and 2018–2022 in its methods. (Waymo)
Study cost: not disclosed.

What the study measured

The study uses liability claims as "a proxy for at-fault collisions," Waymo said. A liability claim is one made against the driver's insurance by someone else, so the measure leans toward crashes where the robotaxi may have been responsible.

The human baselines come from Swiss Re's data on more than 500,000 claims and more than 200 billion miles, according to Waymo. The study also compared Waymo with a narrower group: recent model-year cars (2018–2021 in the paper's abstract, 2018–2022 in its methods) fitted with automatic emergency braking, forward collision warning, lane keeping and blind-spot warning. The gap was a little smaller there, at 86% for property damage and 90% for bodily injury.

Both bodily-injury claims against Waymo were still open, the company said, and are described in the paper.

Waymo presented the claims work as a complement to its separate safety-impact data, which it says shows fewer serious collisions than human drivers over 25 million miles "independent of who was at fault." The Swiss Re study, Waymo said, "provides robust evidence that Waymo had no responsibility in the vast majority of the collisions its vehicles were involved in." That is Waymo's reading of results it helped produce.

The paper's own caveats

The paper lists several limits (Waymo Research). Human mileage was estimated from insurance policy-years, not measured. The garaging zip code of each insured car stood in for the city it drives in. Waymo drove almost entirely on surface streets, while the human benchmark includes freeway miles. Open claims may still close without payment, and more claims may emerge. The authors argue these limits make the results conservative. That is their reading.

How it compares with the last study

The Verge reported that the earlier Waymo–Swiss Re comparison covered 3.8 million fully autonomous miles in California and Arizona. The new dataset is more than six times larger and adds Los Angeles and Austin. The Verge also noted that Waymo is "one of the few" companies willing to publicize its own collisions to make its safety case.

What the companies said

"Auto insurance claims data, traditionally used to assess human driver liability and risk, is a powerful tool in evaluating the safety performance of autonomous vehicles," said Mauricio Peña, Waymo's chief safety officer, in the post.

Ali Shahkarami, global head of P&C Solutions at Swiss Re, said the research "shows how insurance data can help evaluate autonomous vehicle safety." Chris Moore, president of Apollo ibott, a Lloyd's of London syndicate, called a 90% lower claims rate against newer cars "a staggering benchmark." Those are statements from the study's partners and an insurer, not independent reviews of the method.

Why the sample still matters

The Verge pointed out that humans drive close to 100 million miles between fatal crashes. That means "hundreds of millions of miles" from autonomous vehicles may be needed before meaningful comparisons on deaths. At 25.3 million miles, this study can speak to claims, not fatalities. About 40,000 people die on U.S. roads each year, the outlet noted.

What went wrong

Not independent. Waymo co-ran the study and published it, and four of its six authors are Waymo employees. We found no independent check of the results. The paper had been submitted to a journal but was not yet accepted at the cutoff.
Small numbers. The bodily-injury result rests on two claims, both still open.
Liability, not all crashes. Claims capture who was blamed and paid, not every collision.
What we don't know yet: whether the journal accepts the paper, and how the results change as Waymo's miles grow.

Our follow-up question for Waymo

Will Waymo release the claim-level data or the matching method so outside researchers can repeat the Swiss Re comparison?

Later developments (added Oct. 2026)

Waymo's research page now lists the paper as "Do Autonomous Vehicles Outperform Latest-Generation Human-Driven Vehicles? A Comparison to Waymo's Auto Liability Insurance Claims at 25 Million Miles," with six named authors. Its safety research index tags it as peer-reviewed research, but Waymo's own safety hub still described it as "currently under peer review" (checked Oct. 26, 2026). We have not confirmed which journal, if any, published it. (Waymo safety hub) (Waymo Research)
By August 2026, Waymo spokesperson Sandy Karp said its vehicles had driven more than 220 million fully autonomous miles and were involved in an estimated 94% fewer collisions causing serious injuries than human drivers, the Los Angeles Times reported. Waymo's safety hub, with data through June 2026, now shows 271.3 million rider-only miles and 95% fewer serious-injury-or-worse crashes. Those are Waymo's own figures: they count crashes regardless of fault, against a human benchmark built from police-reported crashes in the same cities. (Waymo safety hub) (Los Angeles Times)

Sources

Waymo, The Waymo Team, “New Swiss Re study: Waymo is safer than even the most advanced human-driven vehicles,” Dec. 19, 2024. waymo.com
The Verge, Andrew J. Hawkins, “Waymo still doing better than humans at preventing injuries and property damage,” Dec. 19, 2024. theverge.com
Waymo Research, Di Lillo et al., “Do Autonomous Vehicles Outperform Latest-Generation Human-Driven Vehicles? A Comparison to Waymo's Auto Liability Insurance Claims at 25 Million Miles,” 2024. waymo.com
Waymo, “Waymo safety impact” hub, data through June 2026 (accessed Oct. 26, 2026). waymo.com
Waymo Research, paper PDF with author affiliations. storage.googleapis.com
Los Angeles Times, Caroline Petrow-Cohen, “Waymo wins approval for drastic expansion in L.A., Bay Area regions,” Aug. 14, 2026. latimes.com

Disclosure: [not in copy] Editor placeholder