Skip to content

This piece is a guest contribution, published as part of the Dais’ commitment to being Canada’s platform for bold ideas and better leadership. The views expressed are those of the authors and do not necessarily reflect the positions of the Dais or its staff.

Guest author

Andrew Miller
Co-author of The End of Driving, writes the weekly innovation newsletter Changing Lanes.

Latest from the Dais

Tags

GUEST COMMENTARY

Toronto Should Be Talking About Robotaxis

By Andrew Miller

Illustrations of streets viewed from the top

Waymo, the American robotaxi firm, now carries passengers in fourteen American cities, offering more than 500,000 paid rides per week since March. It is testing in Tokyo and London, and has registered corporate entities in France, Spain and the Netherlands.

We want this technology in Canada too. And that’s not just my view; if we take the Canadian government at its word, it agrees. The federal artificial intelligence (AI) strategy released in June, AI for All, says a trade-dependent country of this size “needs intelligent logistics, autonomous systems, and predictive infrastructure maintenance”. That may not be an explicit endorsement of robotaxis, but there’s no question that a self-driving taxi is the most conspicuous intelligent and automated system a typical Canadian can interact with today. 

The value that such systems offer us is evident. In 2023, 1,964 people were killed on Canadian roads, the highest count in a decade. In most of these, human frailty was part of the cause: unsafe speed in a quarter of them, driver impairment in a fifth, driver distraction in nearly as many. Toronto deserves credit on this file, having halved its road deaths since 2016 to thirty-nine last year. Even one is too many, though, which means we face a moral imperative to solve the problem. A ride-hail driver that never drinks, speeds, gets high, or scrolls their phone is the best, and only, solution available. And such a driver would find a ready market: twenty-eight per cent of Toronto households own no car at all, meaning there are many who prefer not to drive, or who cannot, but who nonetheless occasionally need to travel by automobile.

But Waymo isn’t active in Toronto yet. Why not? 

The firm is surely interested. Last April, Waymo told Toronto’s mayor’s office that it would apply to Ontario’s automated vehicle pilot program. Joining the program would be a small thing, since it authorizes participants to test their vehicles in Ontario and nothing more, so if admitted, Waymo could not carry paying passengers here. But joining would have been a start, allowing the firm to map Toronto’s streets and prepare the ground for revenue service. 

Five months on, as far as the public knows, Waymo hasn’t filed that application, and last month the company named Munich as its next international market. Toronto appears not to be on Waymo’s roadmap. 

It’s reasonable to think that this is because when the firm came to the mayor’s office, the mayor’s response, e-mailed to CBC, was to state that “[t]he mayor will not support Waymo if it costs jobs, drives down wages for other workers, or contributes to precarious work in our city… Waymo must demonstrate this will not happen.”

At least Mayor Chow is on the record on the subject. With the mayoral election coming on 26 October, no other candidate has said anything on the subject at all. 

The best that can be said of this silence is that licensing and regulating automated vehicles is a provincial matter. But it’s Canada’s cities that license vehicles for hire, meaning they have a big stake in the subject as well. Residents should expect candidates for mayor to state, before the vote, what rules they want their provinces to write. 

Toronto has little to work with

Ontario has run its automated-vehicle pilot since 2016, making it the first jurisdiction in Canada to offer one. Over the past decade, fifteen participants have signed up; at the province’s last count, three remained: Gatik, the University of Waterloo and Magna, though Magna ended its Toronto deployment last autumn. None of them may sell a ride, but only test their vehicles on public roads. It’s a low bar, but the highest in the country: British Columbia bans vehicle automation outright, Quebec’s autonomous bus order lapsed in 2023, Saskatchewan’s testing law has been drafted but unproclaimed since 2020, and Nova Scotia rewrote its entire traffic law last year without mentioning driving automation at all.

It’s worth noting that in August 2025 Ontario launched a ten-year program for automated trucks, including a driverless option, that permits hauling of freight: the program conditions say explicitly that automated trucks “may engage in commercial activities during testing.” So Ontario will permit a driverless truck to haul freight in return for money… but won’t permit a robotaxi to carry a person.

Meanwhile, the federal government is ducking the issue. February’s automotive strategy promised to modernize “vehicle safety regulations covering connected vehicle technologies” (note that’s connected, not automated) and punted automation to the then-coming artificial intelligence strategy. That strategy, the aforementioned AI for All, names transportation as a priority sector in a single sentence and commits nothing. Across fifty pages the words vehicle, truck, self-driving and driverless do not appear once.

So a mayoral candidate in Toronto (or Vancouver, or any Canadian city) looking for something to campaign on will find nothing helpful: neither standards, nor timetable expectations, nor funds. Toronto’s AI strategy is not out yet. Candidates will have to use their own judgment about what the city needs, and voters will have to do the same.

What candidates should say

Here’s what voters should expect. They should say that robotaxis are welcome to operate in Toronto, subject to two conditions. 

The first is disclosure. Waymo’s account of its own safety is that, across 220 million driverless miles, its vehicles were in 94 per cent fewer serious-injury crashes and 82 per cent fewer injury crashes than human drivers would have been. In July the Insurance Institute for Highway Safety (IIHS) did its own analysis, and put Waymo in 68 per cent fewer crashes per mile than human drivers, ranging from 76 per cent fewer in Phoenix to four per cent more in Austin. So Waymo is better than human, but how much better? The IIHS concluded that right now data monitors in the USA are not doing a good enough job. And beyond the safety record, almost nothing about robotaxi operations is measured at all: California compels Waymo to file quarterly trip and mileage data, but no jurisdiction requires an operator to report anything about the market it enters. So Ontario should require, as a condition of any licence, that operators publish all data about every collision their vehicles are involved in, with fault recorded, much as California does. Beyond this, Ontario should require operators to build the dataset the province will need in the future, including how many trips are served and how many kilometres the robotaxis travel empty. Separately, it should do its own research into what is happening in the ride-hail market. That data will inform the policies that should be in place before the robotaxis scale.

The second is compensation; offsetting losses for ridehail drivers who lost work. Toronto already charges thirty-five cents on every ride-hail trip. Ontario’s licence could require an additional fee on automated ridehail trips and let the city collect it, directing the proceeds to retraining and income support for ridehail drivers. Portland is debating measures very much like these. Some may think it rich that the city that put up a token resistance to ridehail drivers collapsing the market for regulated taxis will exert itself to help ridehail against the robotaxis, but that’s politics: taxing winners to make losers whole is the bargain that allows innovation to happen in public at all. 

Ontario’s pilot regulation will expire in October 2027, and that regulation required the Minister to evaluate its success before January 2026. If that has happened, it hasn’t been published. That means that, if Toronto had requests to make, now is the time, since Queen’s Park will have to take some steps about driving automation in any case.

The urgency is compounded by our knowledge of what happens if governments fail to prepare. Last month Nevada authorized seven thousand robotaxis in Clark County in a single sitting, without any conditions on published safety data; whatever they might have liked to do, they felt they had to move fast and narrow to accommodate the applicant. Canada’s provinces still have leisure to prepare for what their residents want, and shaped to the sector, not to any particular firm. 

Canada’s cities have an obligation to seize what these vehicles can offer while minimizing their costs. That means that Toronto’s candidates should say what they want from robotaxis, and say it now, loudly enough that Queen’s Park has to respond.