Waymo has announced that its first European Union service will launch in Munich, making Germany its fourth market after the United States, the United Kingdom and Japan.
The company points to more than 20 million completed trips and over 350 million kilometres driven for the public without human control.
But the question facing European cities is no longer whether the technology works. It is this: when something goes wrong, who intervenes?
Europe has not yet written that answer down.
Where robotaxis actually are in Europe
Deployment is moving faster than most regulatory timelines assumed.
Zagreb hosts Europe’s first commercial robotaxi service. Verne, founded by Rimac Group’s Mate Rimac and running Pony.ai technology, began paid operations on 8 April 2026, currently with a safety driver on board.
Uber and Pony.ai announced on 14 August that they will deploy more than 2,000 robotaxis across four additional European cities. The cities were not named.
Madrid is preparing Spain’s first robotaxi pilot through the WeRide and Uber partnership. Zurich is another named city in the same programme.
London is the most crowded field. Waymo has been testing with safety operators since April. Baidu is testing with Lyft and Freenow, with public rides planned for 2027. Uber and Wayve have also announced a service there.
Berlin is deploying autonomous ID. Buzz vehicles through BVG and MOIA under the NoWeL4 project: five vehicles across a 15 square kilometre area in the north-west of the city, serving around 80 stops.
Leuven hosts WeRide’s Robobus under Belgium’s first federal Level 4 test permit. Luxembourg has Stellantis and Pony.ai testing autonomous Peugeot e-Traveller vans. Hamburg has a Volkswagen project running with safety drivers.
Waymo has reportedly established entities in France, the Netherlands, Spain and Germany.
What happened in San Francisco
To understand which rules Europe needs, it helps to look at a single day in the United States.
On 20 December 2025, a power outage knocked out traffic signals across large parts of San Francisco. The city’s 911 dispatch centre soon began receiving calls about driverless vehicles stopped in travel lanes.
According to figures reported from Waymo’s own accounting, the fleet stopped 1,593 times for two minutes or more. Dispatchers called the company 31 times to get vehicles moved.
One caller was left on hold for 53 minutes.
That last detail captures the structural constraint of the current robotaxi model precisely. The vehicle may be driverless, but the system still depends on a human being. And that human is at the end of a support queue.
Emergency response is the sharpest edge
San Francisco is not an isolated case. The documented pattern in the US includes robotaxis blocking ambulances and fire engines, entering active crime scenes, and failing to recognise basic safety signals such as traffic cones and flares.
San Francisco Fire Chief Dean Crispen has said vehicles are picking up and dropping off passengers directly in front of fire stations and ambulance deployment facilities, obstructing emergency access.
The US National Highway Traffic Safety Administration has told driverless vehicle developers that blocking ambulances, firefighters or police is not a rare edge case.
In July, Representative Kevin Mullin introduced the AV Emergency Response Coordination Act at a press conference joined by San Francisco’s mayor and fire chief, citing an unacceptable number of incidents in which autonomous vehicles have interfered with emergency responders.
California’s answer, and why Europe should read it
California has produced a written response, and it is the most directly transferable document available.
Assembly Bill 1777 took effect on 1 July 2026. It does three things.
A 30-second rule. Operators must provide a dedicated emergency line and two-way voice communication allowing emergency officials to reach a remote human operator within 30 seconds. The 53-minute hold in San Francisco is the direct justification for this clause.
Emergency geofencing messages. Authorities can instruct autonomous vehicles to leave a defined area because of an emergency. Rather than calling about vehicles one at a time, a city can clear a zone.
A noncompliance notice. Peace officers can issue a notice of autonomous vehicle noncompliance for a vehicle code or local traffic ordinance violation committed while autonomous technology is engaged.
That third measure creates a working enforcement mechanism where there is no driver to cite. China’s draft Road Traffic Safety Law, submitted for first reading this week, approaches the same gap from a different direction: it names the manufacturer or importer as the party responsible for handling violations committed while autonomous driving is engaged.
Two jurisdictions, one shared realisation. Existing traffic law assumes a human being at the receiving end of a penalty.
Will they reduce traffic? The data says no
The most common argument for robotaxis is that they will ease congestion. The available evidence does not support it.
A 2026 analysis of California Public Utilities Commission data covering roughly 86 million vehicle miles and 14 million Waymo trips found that empty and pickup-related miles remain a substantial share of total robotaxi travel.
A separate 2026 meta-analysis found an average 5.95 percent increase in vehicle miles travelled across the studies it examined. That finding covers automated vehicles broadly rather than forecasting any single robotaxi market, but it is sufficient to discard the assumption that automation reduces traffic on its own.
The mechanism is intuitive enough. A driverless vehicle travels empty to a pickup. It repositions between demand areas. It circulates while waiting, because circulating is often cheaper than parking.
None of this makes robotaxis bad for urban mobility. It makes clear that somebody has to decide, in advance, who bears the cost of a vehicle that moves continuously instead of parking.
Europe’s structural gap
The difference between Europe and the United States is not that European regulation is stricter. It is that European regulation is elsewhere.
The EU-level framework for autonomous vehicles remains in development. The binding constraint sits at member state level, where most national traffic laws still require a human driver at the wheel.
The provisions needing revision are well identified: driverless vehicle licensing, passenger safety, data collection, and the allocation of liability after a crash. Some member states are further along than others, with Germany notable for its early Level 4 legislation.
The United Kingdom is on a separate track outside the EU. Its Automated Vehicles Act replaces the driver with the concept of a user-in-charge and requires autonomous vehicles to demonstrate they are at least as safe as human drivers. Pilots were fast-tracked to spring 2026, with wider rollout expected once the full Act takes effect in the second half of 2027.
The pattern is consistent across the continent: the vehicles are arriving ahead of the rules.
Five questions cities should settle first
The US experience maps out clearly what European cities need to resolve before signing agreements with operators.
Who is reachable in an emergency, and how fast? California said 30 seconds. That is measurable and can be written into a contract.
Is there authority to clear a zone? After a fire, a demonstration or a crash, can a city order vehicles out of an area?
How is a violation processed when there is no driver? California created a noncompliance notice; China names the manufacturer. A third approach is possible, but an approach is necessary.
Will empty miles be reported? Mandating this data is the only way to measure traffic impact, and it is far harder to obtain retrospectively.
How is accessibility guaranteed? Minimum wheelchair-accessible vehicle provision, geographic coverage and transparent denial rates are conditions that are difficult to add after launch.
The real question
Whether robotaxi technology works is no longer seriously contested. Waymo’s 350 million kilometres have largely settled that argument.
What remains open is different: how a driverless vehicle lives inside a city that already has people, rules, fire crews and municipal budgets.
The 53-minute hold in San Francisco was not a failure of the technology. It was a failure of institutional coordination. European cities still have the opportunity to learn that lesson without living through their own 20 December.
Sources
- Euronews, Waymo’s Munich announcement, 26 August 2026
- CNBC and Tech.eu, Uber and Pony.ai European partnership, 14 August 2026
- TechCrunch, Uber autonomous vehicle partnership tracker, August 2026
- Northeast Times, AV Emergency Response Coordination Act, July 2026
- Forbes Technology Council, analysis of the 20 December 2025 San Francisco outage and AB 1777, June 2026
- California Assembly Bill 1777
- Analysis of California Public Utilities Commission robotaxi travel data, 2026
- UK Automated Vehicles Act
- Assessments of EU and member state autonomous vehicle regulation, 2026
