The data gap is itself one of the most important findings
Congress required recurring federal review of rideshare and taxi safety through Sami’s Law. In its July 2026 report, the U.S. Government Accountability Office examined company data, local police records, federal information and safety practices. Its conclusion was not that assaults cannot be measured at all. It was that the available sources do not combine into a complete national dataset.
That matters because public discussions about rideshare safety often jump from one number to a national conclusion. A company report may cover only certain categories. A police database may not reliably identify whether a ride involved Uber, Lyft or a taxi. A federal crime dataset may not contain the trip-level information needed to isolate ride-hailing incidents.
What the companies had publicly reported
GAO reported that two selected rideshare companies publicly disclosed a combined 52 fatal physical assaults and 4,469 reports in the five most serious sexual-assault categories for 2021 and 2022. Those incidents occurred across billions of rides. GAO also described a much larger number of reports falling into other sexual-assault or sexual-misconduct categories from one company.
The denominator is important, but so is the category definition. A reader cannot responsibly compare two companies by pulling one headline number from each unless the covered years, incident definitions and ride counts line up.
The missing years are a major problem for trend claims
As of June 2026, GAO said selected rideshare companies had not publicly released comparable assault counts for 2023 through 2025. That means a simple question such as “Are rideshare assaults becoming more or less common?” did not have a clean current national answer from public company data.
GAO also reported that three rideshare companies declined its request for nonpublic assault data. That does not prove the unreleased data would show a particular trend. It does mean outside researchers and the public are working with an incomplete record.
Police data add detail without solving the national problem
GAO analyzed local records in Los Angeles and Chicago to understand what police data can contribute. The agency identified hundreds of rideshare- or taxi-related assaults in the Los Angeles data it reviewed and more than one hundred in Chicago during the periods studied. In both datasets, those incidents represented less than one percent of all assaults recorded in the relevant jurisdiction and period.
Local records can help describe where incidents happened and how law enforcement classified them. But local systems use different coding practices, and rideshare involvement may appear in narrative text rather than a standardized field. That makes city-to-city comparisons difficult.
Taxi data are even thinner
GAO’s taxi review illustrates the same limitation. One national taxi company provided verified physical- and sexual-assault counts for 2021 through 2025, but GAO cautioned that the selected companies and available records were not nationally representative. There is no single nationwide taxi dataset that functions as a clean control group for Uber and Lyft.
That is why “taxis are safer” and “rideshare is safer” can both be overconfident statements when presented as national facts. The available evidence can answer narrower questions, but the broad comparison is still constrained by inconsistent reporting.
What a careful safety claim should say
The strongest conclusion from the 2026 federal review is not that rideshare is uniquely dangerous or uniquely safe. It is that serious incidents are measurable, companies have added safety tools, and the public data still have meaningful gaps.
For riders, drivers, lawyers and policymakers, that incomplete picture has practical consequences. It affects how trends are evaluated, how platform claims are compared, and how confidently anyone can say that one transportation mode or company is safer than another.
Sources
