On Friday, Waymo announced that Toronto will be its first Canadian city. Its white Jaguar SUVs will begin driving Toronto streets in the coming days with a trained specialist at the wheel, building the high-definition maps the system runs on. Driverless testing follows. The company intends to carry its first paying riders in 2028. Element Fleet Management, headquartered in Toronto, will run charging, cleaning and maintenance, and the two companies say they will create hundreds of roles over time.
The announcement gives no fleet size, fare, or date for driverless testing, and no plan for Pearson or the Island airport. The testing area Waymo described to The Logic covers the downtown core and waterfront, midtown, and parts of the east and west ends. The company says its all-weather driver will adapt to Toronto’s conditions and gives no operating limits for snow, ice or visibility. Waymo has entered more than a dozen American cities the same way, mapping first and setting terms later.
The difference in Toronto is the legal ground. Every American city where Waymo carries riders sits in a state whose law permitted commercial driverless service before the first paid ride. In Ontario, no such law exists, and the testing permission Waymo is working under expires before its launch date.
I wrote in March about what robotaxis would mean for Toronto. This post is about the narrower question Friday’s announcement raised. Waymo has said when it intends to open for business, and the rules it will open under have to be written in the meantime. Uber launched in Toronto in 2012 and the city passed a bylaw that fit it in 2016. This time the timeline is public in advance.
Autonomous Taxis in Toronto
I remember the first time I tried Uber. My home at the time was a long walk from the nearest transit stop, it was winter, and I was going to a alcohol-infused event. So instead of braving the cold or…
The rules that exist today
The province decides whether an automated vehicle may drive on a public road. The city licenses vehicles for hire. Waymo needs both, and neither covers what it plans to sell.
Ontario’s permission is a pilot. The Automated Vehicle Pilot Program began in 2016, the first in Canada. Participants must own their vehicles, carry at least $5 million in liability coverage, report every collision within ten days, and file an annual report on testing speeds, road types and disengagements. Since 2019 the program has allowed testing with no one in the driver’s seat, if the participant files its operational design domain, meaning the environments and limits the system is built for, and a plan for how the vehicle behaves around work zones and first responders. No participant may sell a ride. The regulation now runs to October 13, 2027.
The province has written the other kind of rule for freight. Its ten-year program for automated commercial trucks, launched in August 2025, states that participants may engage in commercial activity during testing, and Gatik already runs autonomous box trucks for Loblaw in Toronto under it. Andrew Miller, who writes the excellent Substack Changing Lanes, writing for the Dais at Toronto Metropolitan University, points out the result: Ontario permits a driverless truck to haul freight for money and does not permit a robotaxi to carry a person.
The city has less to work with. Toronto has no bylaw that mentions automated vehicles. Its Automated Vehicles Tactical Plan covered 2019 to 2021 and has not been replaced. Under the provincial pilot the city receives notice of where testing will happen and asks participants, as a courtesy, to share their work-zone and first-responder plans.
The city does hold the vehicle-for-hire licence. Uber, Lyft and every other ride-hail company operate in Toronto under a Private Transportation Company licence that costs $24,424 to apply for and comes with a data-sharing agreement, insurance minimums, driver screening, and two per-trip charges: $0.35 to the city and $0.10 to the Accessibility Fund. The licence was written for companies that connect passengers with drivers. Extending it to driverless operators is within council’s control.
Washington, D.C. shows what happens when the gap stays open. Waymo has tested there with safety drivers for two years while the District’s council works on a bill to authorize commercial driverless service. Labour groups and ride-hail drivers testified against the bill this year, and Waymo has no launch date.
The jobs question
In April, when Waymo first told City Hall it would apply to the provincial pilot, the mayor’s office said the mayor would not support the company if it cost jobs, drove down wages or added to precarious work, and that Waymo must demonstrate this will not happen. On Friday the mayor’s stated concern was safety, and the point that the enabling legislation is provincial.
The April condition deserves a direct answer. A driverless taxi service reduces the work available to ride-hail drivers, because removing the driver is what the service sells. The hundreds of roles Waymo and Element describe are real jobs, and they will not replace the far larger number of people who drive for Uber and Lyft in Toronto, whose median earnings after expenses the city’s 2024 review put below the minimum wage. No operator can demonstrate that the displacement will not happen. A condition that cannot be met functions as a refusal, and it leaves the decision to the province.
The workable version of the same concern is compensation. Toronto already collects $0.35 on every ride-hail trip. A higher per-trip charge on driverless trips, collected by the city and directed to income support and retraining for displaced drivers, is a condition an operator can accept and a council can enforce. Miller makes this case, and Portland is debating a similar measure. Toronto did little for taxi licence holders when ride-hail collapsed their market a decade ago. Doing better this time costs the city nothing it does not raise from the new service.
The trade-off is between a service that is safer than what it replaces and a loss of flexible work for people who have few alternatives. Both halves are true. The policy that honours both is to let the service in and tax it to compensate the people it displaces.
What the province should require
Queen’s Park has to act on the pilot before October 2027 in any case. It can extend a testing regime for a third time, which leaves Waymo mapping a city it cannot serve, or replace it with a framework for commercial driverless service. The second matches what is now happening on Toronto’s streets, and the conditions attached to it are the province’s main leverage.
The first condition is disclosure. Waymo’s own account is that across more than 400 million fully autonomous kilometres its vehicles were in 95 per cent fewer crashes involving serious injury than human drivers in the same cities. The Insurance Institute for Highway Safety’s independent analysis, published in July, put the reduction in police-reportable crashes at 68 per cent across four cities, from 76 per cent fewer in Phoenix to 4 per cent more in Austin on a small sample. Both findings point the same way, and the spread shows how much depends on who counts and where. The IIHS also concluded that the current American reporting system cannot support monitoring at scale. California requires operators to file quarterly trip and mileage data; no jurisdiction requires reporting on the market an operator enters. Ontario should require, as a condition of a commercial licence, publication of every collision with fault recorded, trips served, kilometres driven empty, and the operating limits the system is cleared for. The province already collects collision and disengagement reports from pilot participants each year and has not published them.
The second condition is the per-trip charge on driverless service, set provincially so it applies across the GTA and collected by the host municipality so the money stays where the displacement happens.
Both conditions should be written for the sector, not for one company. Reece Martin at The Transit Brief makes this point about every policy in his post: Waymo is the first operator to arrive and will not be the last, and a clear operating environment is more likely to attract competitors than to deter them. Nevada authorized 7,000 robotaxis in Clark County in a single sitting last month with no published safety-data conditions. Ontario has a year before its own deadline and a clear statement of the applicant’s intentions.
What the city should require
Council does not need to wait for the province to settle the parts it controls. Three can be drafted now as amendments to the vehicle-for-hire bylaw.
A licence category for driverless operators. The Private Transportation Company licence assumes a driver to screen, insure and pay fees for. A driverless category would carry the same data-sharing agreement, extended to the fields the province should also require, the same insurance floor or higher, and the per-trip charges. Writing it now means the terms exist before the first application rather than being negotiated against one.
A per-kilometre charge on empty and central kilometres. Martin proposes a congestion charge applied only to driverless vehicles. It avoids the objection that has stopped road pricing in Toronto so far, which is that it falls on people who already drive. Driverless vehicles carry precise GPS, so the charge can be calculated on distance travelled inside a zone during peak hours and on kilometres driven with no passenger aboard. No current driver pays more. The operator gains a faster road. The city gains a revenue stream that scales with the service and can fund the surface-transit priority the service will otherwise compete against. The streetcar network is the most exposed part of the system. The city’s 2024 transportation impact study found that between 41 and 61 per cent of ride-hail users would have taken transit had ride-hail not been available, and a cheaper, more comfortable ride-hail will pull harder on the same trips.
Curb rules. A driverless vehicle picks up and drops off wherever its software judges safe, and the record from San Francisco and Phoenix shows that construction zones, bike lanes and emergency scenes are where that judgment fails. Toronto should designate pick-up and drop-off zones on the streets where volumes will be highest, prohibit stopping in a bike lane or at a transit stop outside an emergency, and make the work-zone and first-responder plan a licence requirement rather than a courtesy. These rules apply to ride-hail generally. The arrival of a new operator is the point at which the existing ones are least able to object.
None of this requires council to take a position for or against robotaxis. It requires council to set its licence terms before the province writes the rule the city will have to work within.
Winter
The first reaction many Torontonians had on Friday was the snow. Mathew Growden at The Change Optimist asked in his headline whether the Waymo driver is ready for it, and his answer is that Waymo has not said. The company ran tens of thousands of miles of winter testing in Upstate New York and Michigan’s Upper Peninsula last October, its sixth-generation hardware heats and cleans its sensors, and its software slows the vehicle on slippery surfaces. What it has not published, for Toronto or anywhere else, is the limit: the snowfall rate, the visibility, the ice condition at which the service pauses.
Martin’s counterpoint also holds. Toronto is not the snowiest large city in Canada. Waymo has carried riders in Denver since August, is testing without a driver in Detroit, and began mapping Chicago and New York this year. Two years is a long runway by Waymo’s standards, and winter is the likeliest reason for it. The company will see two of them before it sells a ride.
The province already requires a driverless testing applicant to file its operational design domain. The condition that matters for a commercial licence is that the operator publish it and update it as the limits change. Riders can then judge whether a service that pauses in a storm is worth relying on, and the city can plan around the days it will not run.
What next?
The election is on October 26. Licensing automated vehicles is a provincial matter and vehicle-for-hire licensing is a municipal one, so whoever is mayor in November will spend the following year either shaping the provincial rule or receiving it. Miller’s test for candidates is a fair one: say whether robotaxis are welcome, and on what terms, before the vote.
The province should, before the pilot expires in October 2027:
Replace the testing pilot with a commercial framework for driverless passenger service, written for the sector rather than for one applicant.
Require, as a licence condition, publication of collision data with fault recorded, trips served, empty kilometres, and operating limits for weather and visibility.
Authorize a per-trip charge on driverless service that municipalities collect and direct to transition support for displaced drivers.
The city should, in the first half of the new term:
Add a driverless operator category to the vehicle-for-hire bylaw, with the data-sharing agreement extended to the same fields.
Adopt a per-kilometre charge on driverless vehicles for peak-hour central kilometres and for kilometres driven empty, with the proceeds committed to surface-transit priority.
Designate pick-up and drop-off zones on high-volume streets, prohibit stopping in bike lanes and at transit stops, and make the first-responder interaction plan a licence requirement.
Waymo has given Toronto the dates. The rules it will operate under do not exist yet, and both governments that will write them have a reason to act within the year. Two years is enough time for the city to state its terms. If it does not, the province will write the rule without them.