Four things the data shows
Every number here is computed from the published dataset at build time and links to the page where you can check it. If a figure below disagrees with the map, the map is right and this page is broken — tell us.
Most of what Canada's ALPR map calls surveillance is the 407
244 of 931 nodes tagged as ALPR surveillance in OpenStreetMap — 26% — are tolling infrastructure, not policing.
Of those, 244 are the 407's own cameras: manufactured by RTX (Raytheon, which built the 407 ETR tolling system) and operated by the Ontario Ministry of Transportation. Every one of them falls inside the 407 corridor. They read plates because billing by plate is the service — you entered a toll road.
Existing crowdsourced maps show these as generic surveillance cameras, which inflates the apparent scale of police surveillance by a wide margin. Readback classifies them separately and turns the layer off by default, because a toll you chose to pay is not a dragnet on your commute.
How to check it: filter /data/cameras.json for
source: "alpr_tagged" and count category: "toll". The
classification rule is in build-cameras.mjs,
and each node links to its OpenStreetMap record.
See it on the map, toll layer on → · The MTO cameras, listed → · The data →
First identified 2026-09-01, when the same query returned 244 RTX/MTO nodes out of 618 ALPR-tagged — 40%. The dataset has grown since; the finding held. Both figures come from the same rule applied to OpenStreetMap on different days.
The scanning that matters most is on no map at all
13 Canadian police services and municipalities run vehicle-mounted ALPR across roughly 783 equipped vehicles.
A dot map can only show cameras bolted to poles. But most Canadian ALPR capacity is mobile: cruisers scanning continuously while they drive. Those have no fixed location to map, so every existing ALPR map — ours included, until we added a capacity layer — structurally undercounts the thing it claims to measure.
This is why we render fleets as capacity circles rather than pretending not to know. It is also why a "how many cameras are there" framing is the wrong question.
Retention is a setting, not a standard
24 hours in Peel Region. 7 days in Toronto. Same province, same technology, neighbouring forces.
Across the jurisdictions we track, how long a plate read of an innocent driver is kept runs from 3 minutes to indefinitely. Nothing about the hardware requires any of these numbers. Retention is an administrator's dropdown, and in most of Canada no statute constrains it.
The consequence is the site's central claim: every safeguard on these systems is a setting, and nobody outside each service checks what the settings are. That is not an accusation that anyone is lying. It is an observation that nobody is in a position to know.
The full comparison, with sources → · The requests we filed to find out →
Some plate readers also listen for your phone
Two vendors sell ALPR units that additionally capture Bluetooth and Wi-Fi identifiers from passing vehicles. No Canadian retention rule mentions device signals at all.
Jenoptik's TraffiCatch has been deployed in US border counties since around 2019. Leonardo markets comparable capability. More importantly, the radios needed to do this are already present in hardware from several major vendors — which means the capability is, again, a setting rather than a purchase.
Retention rules, privacy impact assessments and oversight frameworks in Canada are written about plates. A device identifier is not a plate, and the rules are silent.
What is mapped, in total
Press enquiries, data, or verification: [email protected]