Uber’s promise of a vetted driver network has collided with a patchwork of rules, internal choices and technical limits — and the result, according to a months‑long review, is that people with serious violent convictions have slipped back onto the road.
At issue are two policy choices that look deceptively simple but have big consequences. Uber disqualifies applicants for a handful of “top‑tier” crimes — murder, sexual assault, kidnapping and terrorism. Beyond that, in many states the company treats older violent convictions differently: in roughly 22 states, convictions for assault, stalking, child abuse and similar offences can be approved if they’re more than seven years old. Add a second wrinkle — in 35 states the company’s checks focus on places an applicant lived in the last seven years rather than searching all jurisdictions — and the system has obvious gaps.
A policy with holes
The seven‑year cutoff is presented inside the company as a balance between safety and giving people a second chance. Uber’s safety lead for the Americas has described the timeframe as rooted in research suggesting recidivism declines over time. But safety advocates and some criminal‑justice researchers say the blanket rule ignores the different risk profiles of violent crimes and the patchiness of public records.
That geographic limitation is especially consequential. If an applicant lived in three different counties or states in the past decade and Uber’s screening only touches the two most recent, earlier convictions can go undetected. And because background‑check techniques vary — from name‑and‑date searches to fingerprint‑based checks — the odds of catching older or out‑of‑jurisdiction convictions fluctuate widely.
Uber executives, per internal documents, debated whether to expand exclusions or to invest in more exhaustive screening. A 2015 memo flagged alternatives such as fingerprint checks and in‑person interviews but also noted practicality and cost. At least one former internal email described the company’s background policies as “a bare minimum.”
When gaps become danger
The abstract policy question has landed in courts and on police reports. The review compiled several instances where drivers who had past violent convictions were later accused by passengers of sexual assault; two of those cases resulted in criminal convictions. Separately, company records show reports of sexual assault or misconduct occurring at a striking cadence: in the years reviewed, Uber’s internal tallies indicated a report roughly every eight minutes between 2017 and 2022. The company insists most reports are for less serious behavior and that 99.9 percent of rides happen without incident, but even a small percentage of millions of rides is meaningful in human terms.
Some states have acted. A 2017 audit in Massachusetts found thousands of drivers who should not have been on the road under stricter rules — more than 8,000 drivers were banned after an audit. Competing platforms take different stances: one major rival bars drivers with violent felony convictions regardless of when they occurred, reflecting a stricter, lifetime‑exclusion approach for certain offenses.
Uber’s defense and the counterarguments
Uber’s public explanation is twofold. First, the company argues that long‑ago convictions are a poor predictor of future violent behavior and that lifetime bans prevent people who have served their sentences from rebuilding. Second, Uber has pushed back on some screening methods on technical and fairness grounds: company statements have called fingerprint checks inaccurate, ineffective and potentially discriminatory, and warned that in‑person interviews could introduce bias.
Critics counter that the environments in which people with violent convictions are already excluded — daycare, eldercare, schools — reflect a societal judgment about risk that should apply to platforms that put drivers into close, often isolated contact with vulnerable passengers. They also point out that public‑records searches can miss convictions in jurisdictions that don’t easily surface records online, so relying on limited windows of residence makes the odds of false negatives—missed red flags—much higher.
What could (and should) change
There’s no single fix, but several options are being discussed across the industry and by safety experts:
- Expand the scope of checks for violent crimes and eliminate state‑by‑state inconsistencies where public safety is concerned.
- Use more robust identity verification and continuous monitoring rather than a one‑time check. Technology can help here — better integration between booking platforms and maps/navigation tools, or automated reporting pathways, can shorten the time between an incident and remediation. (Some of those capabilities align with advances in services like Google’s AI Mode for agentic bookings and the conversational navigation copilots rolling into mapping apps.)
- Improve data sharing with law enforcement while protecting civil liberties, and invest in better human review of potentially problematic records.
- Reconsider whether certain violent offenses should be treated the same as lower‑risk nonviolent convictions in safety‑critical contexts.
There’s also a role for product design. Features like emergency buttons, mandatory ride‑sharing of trip data with trusted contacts, real‑time monitoring of route deviations and recorded audio snapshots (with strict privacy limits) can reduce harm once a ride begins. Navigation and in‑ride systems are getting smarter — as a conversational AI copilot for maps shows, navigation tools are evolving and could be adapted to flag anomalies in routes or timings that suggest a problem.
A company balancing growth and risk
Underlying the debate is a business pressure that most tech platforms know well: growth and onboarding speed are expensive, and adding layers of verification slows signups and raises costs. Internal discussions show Uber explicitly weighed those tradeoffs, considering whether to focus messaging on “safety initiatives” that cost less than the most exhaustive background work.
That calculus is what prompts the heated arguments today. For passengers who have been harmed, the math feels cold and transactional. For policymakers and safety researchers, the question is whether a technology‑driven industry should accept patchy vetting because it’s cheap and fast.
Whatever the next steps, the problem is not insoluble. It will take a combination of clearer policy — particularly around violent offences — smarter use of technology, and transparency about how and why decisions are made. Until then, millions of rides continue to be made each day under rules that some argue are too loose for the risk involved.