Industries / Transport and vehicles / Ride-share and taxis
Safety
Safety covers protecting riders and drivers from assault, robbery and crashes during a hired ride.
4
Milestones achieved, from 2013 to 2026.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
2013
Cameras cut cab killings
NIOSH researchers compared 26 US cities over 15 years. Cities whose taxis carried security cameras had a driver homicide rate three times lower than cities with neither cameras nor partitions, and seven times lower than before the cameras went in. The study was published in the American Journal of Preventive Medicine and relied on newspaper clippings.
Credit
NIOSH Division of Safety Research
2019
A company counts its assaults
In December 2019 Uber published its first US safety report: 3,045 reported sexual assaults, 9 murders and 58 crash deaths on US Uber trips in 2018. Finding "no uniform industry standard" for classifying such reports, Uber worked with the National Sexual Violence Resource Center and the Urban Institute on a classification system (a taxonomy) and released it for other firms to use.
Credit
Uber; National Sexual Violence Resource Center; Urban Institute
2021
Rivals share banned drivers
On March 11, 2021, Uber and Lyft started the Industry Sharing Safety Program. Through it they share basic information on drivers removed for the most serious sexual assaults and for physical assaults that killed someone, so those drivers cannot simply move to the other app. The background-check firm HireRight runs the exchange, and other companies may join.
Credit
Uber; Lyft; HireRight
2026
Full categories disclosed
On May 22, 2026, Lyft published 2017–2022 counts for every category of reported sexual assault and misconduct in the RALIANCE taxonomy, the shared classification system, each also given as a share of completed US rides. Its earlier reports covered five categories.
Credit
Lyft; RALIANCE
Watch for
Women riders choose women drivers
Uber extended its "Women Preferences" option across the US. Women riders can ask for a woman driver, which improves the odds of one but does not guarantee it. Uber says the feature serves safety. Two California drivers have sued, arguing it breaks the state's ban on sex discrimination by businesses.
Achieved whenan outside party publishes a measurement of whether reported incidents change for riders who use the option.
Claimed by
Uber, March 2026
Open
01
One table for every company
A single public table of reported incidents per million rides, by year and taxonomy category, for every company that publishes figures. It starts with Uber and Lyft, since both use the same taxonomy, and notes where definitions differ.
The briefCredit · unclaimed02
Update the camera study
A new analysis of whether in-car cameras protect drivers, adding years after 2010 and app drivers, who were not part of the NIOSH study.
The briefCredit · unclaimed
Open 01 · the brief
One table for every company
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A single public table of reported incidents per million rides, by year and taxonomy category, for every company that publishes figures. It starts with Uber and Lyft, since both use the same taxonomy, and notes where definitions differ.
Why it is still open
Each company publishes its own report in its own format. Lyft's 2026 update gives each category as a share of rides but does not compare itself with anyone. This research found no independent table that puts the companies side by side.
Built with what exists
The companies' published safety reports; an AI model that reads long PDFs to extract the tables, checked by hand (the site's note "Seven more flagged" covers using such a model to review long documents); a spreadsheet.
First step
Put Uber's and Lyft's category lists side by side and mark which match exactly.
Done looks like
A published table that covers every year and category both companies report, with each number linked to its page in the source report and a list of every definition that differs.
Open 02 · the brief
Update the camera study
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A new analysis of whether in-car cameras protect drivers, adding years after 2010 and app drivers, who were not part of the NIOSH study.
Why it is still open
The NIOSH study relied only on newspaper clippings, and NIOSH said more research was needed. This research found no update that includes app drivers.
Built with what exists
The US Bureau of Labor Statistics' Census of Fatal Occupational Injuries; city taxi camera rules found through public records; news archives; R or Python for the statistics.
First step
List which of the original 26 cities still require cameras today and which have added or dropped the rule.
Done looks like
A preprint covering at least 26 cities over at least 10 more years, with data and code published, that reports homicide rates for camera, partition and control cities and states whether the 2013 result holds.