Autonomous taxi safety and regulation in 2026
Autonomous taxi safety in 2026 stands on two legs: measured crash data that now favors the machines, and a regulatory framework that finally has global form. On the data side, the largest public dataset shows Waymo's driverless fleet with 92% fewer serious-injury crashes than comparable human drivers across 56.7 million rider-only miles, including a 0.41 versus 2.80 any-injury crash rate per million miles. On the rules side, the UNECE working party adopted a draft global technical regulation for automated driving systems in January 2026, roughly ten years in the making, built on a safety-case approach that makes operators prove safe behavior with evidence across the vehicle's entire life. Together the two legs define what any country, including small markets, will ask of a robotaxi fleet: show the data, document the system, keep proving it.
What does the safety data actually say?
The Waymo dataset is the reference because it is large, public and peer-reviewed. Key comparisons at 56.7 million rider-only miles, against human benchmarks on the same road types:
| Metric (per million miles) | Waymo | Human benchmark | Reduction |
|---|---|---|---|
| Any-injury crashes | 0.41 | 2.80 | 85% |
| Serious-injury or worse | 0.02 | 0.22 | 92% |
| Pedestrian injury crashes | vs benchmark | 92% | |
| Cyclist crashes | vs benchmark | 82% | |
The numbers come from Waymo's safety impact data and the peer-reviewed comparison study. Machines lose the crashes humans cause by fatigue, distraction, alcohol and impatience, which are the majority. What machines add, hesitation in ambiguous scenes, wrong calls on rare layouts, gets handled by conservative driving plus remote assistance.
What does the UN global regulation change?
Until 2026, every jurisdiction improvised: city permits in the US, national pilots in China, patchwork in Europe. The UNECE draft global technical regulation gives signatory countries one framework. Its core is the safety case: the manufacturer and operator must present a structured, evidence-based argument that the system is safe, backed by a safety management system covering development, deployment and operation, with independent audits. It was adopted by the GRVA working party in January 2026 and submitted to the World Forum in June 2026.
For fleet operators the practical consequence is documentation. Trip logs, incident records, intervention histories and maintenance trails stop being nice-to-have; they are the safety case's raw material, produced continuously by the fleet management platform.
How do the EU and US approaches differ?
The EU runs type approval: a regulator tests and approves the vehicle type before it may be sold or operated, and the European Commission is preparing targeted type-approval changes with regulatory sandboxes and automated-driving corridors from 2026. The US lets manufacturers self-certify compliance, which is why deployment ran ahead there: Tesla could take safety monitors out of its Austin robotaxis under state-level rules while European fleets still waited for approvals. Countries outside both blocs, Georgia included, typically adapt the UNECE framework, which is the subject of robotaxis in Georgia.
What does an operator have to prove day to day?
- Operational design domain: where the fleet may drive, in what weather, at what speeds, and how the system behaves at the boundary.
- Incident response: what happens after a crash or a stall, including data handover to authorities.
- Human oversight: staffing and procedures of the remote assistance desk.
- Maintenance and cybersecurity: sensor calibration schedules, software update controls, protection of the vehicle link.
Every item resolves to records, and fleets that generate them automatically clear audits at a fraction of the effort. Scale operators already run this way; the volume numbers in the global robotaxi market exist because the compliance machinery underneath them holds.
Where aiTAXI fits
aiTAXI is a robotaxi fleet management platform by aiNOW (Tbilisi, Georgia), and compliance is one of its five modules: trip logs, incident reports and audit files generated from live fleet data, aligned with the safety-case logic of the UNECE framework. The robotaxi fleet management platform is in early access, with a pilot program open for Georgian taxi companies that want the documentation layer in place before regulation arrives.
FAQ
Are driverless taxis safer than human drivers?
On the largest published dataset, yes by a wide margin: 92% fewer serious-injury crashes and 85% fewer injury crashes of any kind across 56.7 million miles, confirmed in peer review.
What is a safety case?
A structured, evidence-backed argument that a system is safe enough to operate, covering design, testing, deployment and monitoring. The UN's 2026 regulation makes it the core requirement for automated driving.
Who approves a robotaxi service in a specific city?
It depends on the jurisdiction: state or city permits in the US, type approval plus national rules in the EU, and adapted international frameworks elsewhere. In every case the operator must show documented safe operation.
Does a small country need its own AV law?
It needs a national decision, but not a from-scratch framework. The UNECE global technical regulation exists precisely so smaller jurisdictions can adopt a ready, internationally harmonized rulebook.