# Verne’s driverless Zagreb tests stop short of an open service Canonical URL: https://www.theroboteconomy.news/articles/verne-zagreb-driverless-test-commercial-boundary/ Publisher: The Robot Economy Editorial byline: Mara Voss, Deployment Desk Published: 2026-09-17T13:50:37.723Z Updated: 2026-09-17T13:50:37.723Z Editorial information: https://www.theroboteconomy.news/about/#editorial-desks Invited passengers can travel without an onboard operator. The evidence separates that trial from the supervised service already available through Uber. ## Summary Verne and Pony.ai announced invited driverless passenger tests on September 10 in Zagreb. Comparing their records with Uber’s August launch shows that test access, public booking and accumulated fleet mileage describe different stages. A broad commercial service without an onboard operator remains a future milestone in the reviewed records. ## Key facts - Verne announced invited passenger tests without onboard personnel on September 10, 2026. - The test route spans 22 kilometers; it is not the same measure as the commercial service’s area. - Uber’s August 19 launch explicitly retained a licensed operator behind the wheel. - Pony.ai’s reported 200,000 kilometers cover the service since April, not a disclosed driverless-test total. ## Lead image Image: https://www.theroboteconomy.news/images/articles/verne-zagreb-driverless-test-commercial-boundary.webp Description: People beside an open rear door of a white Verne robotaxi in a company-supplied Zagreb press photograph. Credit: Verne service-vehicle press photograph; not a photograph of the September driverless trial. Photo: Verne; cropped. ## Report ### Two services, different access Verne and Pony.ai announced on September 10 that invited passengers had begun taking robotaxi test rides in Zagreb without an operator or other personnel onboard. The route runs for 22 kilometers between Verne’s headquarters, a business district and the airport. It is a real change in who sits inside the vehicle, but the announcement describes testing toward a future commercial driverless service. Sources: [s1]() This week’s coverage preserves an important boundary. TechTarget’s September 14 report describes the invited driverless program as running alongside the commercial service, which continues to carry human safety monitors. That distinction is easy to lose in a headline about driverless taxis reaching Europe: a passenger test and a generally bookable service are different products, even when they share operators and technology. Sources: [s4]() Uber’s August 19 launch record provides the earlier baseline. It opened autonomous bookings in selected parts of Zagreb and explicitly retained a licensed operator behind the wheel. Riders could request UberX or Comfort and receive an autonomous vehicle when one was available. The announcement placed fully driverless operation in a later phase, rather than promising that every booking would already be unattended. Sources: [s3]() For riders, this comparison makes access the practical question. An app listing proves that a ride can be requested under the conditions offered there. It does not, by itself, establish eligibility for the separate invited trial. For other fleet operators evaluating Zagreb as a model, the useful next milestone is a documented change in the public booking conditions, not simply another announcement that tests are expanding. Sources: [s3]() This access boundary also appears in Verne’s current service terms. Joining the waiting list does not guarantee access; the company does not promise vehicle availability or acceptance of every ride request. Those terms describe the service generally, rather than announcing a new release of the driverless trial. A waiting-list registration or an app installation therefore cannot establish a passenger’s eligibility for unattended testing. Sources: [s5]() ### The mileage denominator matters Pony.ai reported more than 200,000 kilometers, several thousand customer rides and a passenger rating of 4.7 out of five for the service operating since April. Its September release separately described the move from commercial operation with an onboard operator to driverless testing. The accumulated service figures therefore cannot be read as disclosed results for the new unattended trial. Sources: [s2]() That is a measurement boundary, not evidence that the trial is unsuccessful. The existing figures show reported operating experience across the earlier service period. A decision about unattended expansion needs its own denominator: how many driverless trips were attempted, how many completed, and what support was required. Combining the two operating stages would obscure the question the new phase is meant to answer. Sources: [s2]() ### Removing the seat is not removing operations Verne says Pony.ai’s system performs the driving task while its remote team monitors vehicles without remotely driving them. It also assigns itself fleet operations and the passenger experience. Development of Verne’s purpose-built vehicle continues separately from this validation program. The trial should therefore not be treated as proof that the company’s separate vehicle project has entered commercial service. Sources: [s1]() The operator’s work remains consequential even as the onboard role changes. Remote supervision, the passenger interface and fleet management must function together. Verne’s account establishes those responsibilities, but it does not quantify staffing per vehicle or support demand. Removing an onboard operator creates a different operating arrangement; the source does not establish how much total human work that arrangement saves. Sources: [s1]() The parallel-service distinction also provides a fair reading of the achievement. Invited passengers travelling without an onboard monitor represent progress beyond supervised rides. Keeping commercial rides supervised while that phase is tested does not negate the technical step. It does mean that statements about broad public availability need their own evidence, rather than borrowing the meaning of a narrower trial. Sources: [s4]() ### What would change the conclusion The next useful disclosure would separate driverless trial activity from the cumulative service figures and identify when unattended rides become a public product. A dated access change, accompanied by a defined operating area and trial-specific completion and assistance data, would make the commercial transition easier to assess. Until then, the records support a meaningful pilot milestone with a remaining service-delivery test. Sources: [s2]() ## Conclusion Compares August public-booking conditions, September invited-test access and the April-to-date mileage denominator to identify the remaining commercial-service evidence. Separates onboard-driver removal from total operating labor and from the separate purpose-built vehicle program. ## Disclosures Sponsored: no Desk bylines, AI assistance and editorial policies: https://www.theroboteconomy.news/about/ ## Sources - s1: [Verne and Pony.ai begin fully driverless test rides with passengers on public roads in Zagreb]() — Verne; company-announcement; 2026-09-10T00:00:00.000Z - s2: [PONY AI Inc. and Verne Kick Off Fully Driverless Robotaxi Test Rides in Zagreb]() — Pony.ai; company-announcement; 2026-09-10T09:00:00.000Z - s3: [Uber Launches Autonomous Rides in Europe]() — Uber; company-announcement; 2026-08-19T00:00:00.000Z - s4: [Fully driverless robotaxis now operating in Europe]() — TechTarget; independent-reporting; 2026-09-14T00:00:00.000Z - s5: [Verne Terms of Service]() — Verne; official-record; 2026-09-17T13:44:59.671Z