Starting in December 2025, up to 50 Edmonton Police Service officers began wearing body cameras equipped with live facial recognition technology. The cameras, supplied through Axon and powered by Israeli firm Corsight AI, scan faces in the officer’s field of view and compare them against a watchlist of roughly 7,000 individuals flagged as high-risk. Matches are not delivered to the officer in the field. They are logged and reviewed back at the station. The pilot is the first of its kind in Canada.
That is a significant policing experiment. It deserves a serious public conversation, which is precisely what has not happened.
No Approval, No Problem, Says EPS
Alberta’s Information and Privacy Commissioner, Diane McLeod, stated clearly that she had not approved the trial and had considerable concerns about the technology, particularly around accuracy. Edmonton Police responded that they are not legally required to obtain the commissioner’s approval before proceeding with a proof-of-concept. Both statements can be simultaneously true. They cannot both be reassuring.
Documents obtained by CBC News revealed details the police did not volunteer publicly, including data-sharing language in the vendor contract that could allow sensitive information to flow beyond EPS and concerns about the reliability of the underlying model. There was also a reported critical-fault system outage during the pilot, and as of early 2026 the Edmonton Police Commission and the chief’s committee were still reviewing results before deciding whether to expand, continue, or end the program.
What the Watchlist Actually Is
The categories of individuals on the 7,000-person watchlist are themselves worth scrutiny. The flags include people classified as violent or assaultive, armed and dangerous, escape risk, and high-risk offender. These are not arbitrary designations. But a watchlist is a static document applied against a dynamic world. People on it may have resolved their legal situations. People not on it may present a genuine risk. The camera cannot know either way, and the officer in the field is told nothing in real time, which raises the question of what the system is actually catching and when.
Facial recognition technology has a documented pattern of higher error rates for darker-skinned faces and women. Corsight AI has disputed some of the published accuracy figures, and independent verification of its specific performance on EPS’s watchlist has not been made public. That is a problem. A technology whose error distribution is unknown, deployed in active policing, with no real-time officer notification, should not be described as a straightforward safety tool.
Public Safety and Civil Liberties Are Not Opposites
Supporting effective policing and demanding clear rules for biometric surveillance are not competing positions. They are complementary ones. Officers benefit from tools that actually work. The public benefits from rules that prevent those tools from being misused. The problem in Edmonton is that the pilot launched before the rules existed, and the police are now in the position of defending a deployment that preceded any democratic or regulatory authorisation.
Alberta communities are already wrestling with the implications of data infrastructure, and facial recognition represents one of the sharper edges of that broader question. The province needs a legislative framework for biometric policing before any system moves from pilot to permanent. That framework should answer basic questions: who maintains the watchlist, how is it audited, what is the error-rate threshold below which a technology cannot be deployed, and what happens when a match turns out to be wrong?
Edmonton Is Setting a Precedent Whether It Wants To or Not
Because this is the first Canadian police deployment of bodycam facial recognition, what Edmonton’s police commission decides next will be watched by every other service in the country. If the program continues without legislative guardrails, the message is that Canadian police can deploy live biometric identification against a database of thousands of people with minimal prior authorisation. Ottawa has been slow on AI governance broadly, and that gap falls to provinces to fill.
Alberta’s legislature should not wait for this pilot to expand before establishing the rules. Legislation that defines the permissible scope of police biometric tools, mandates independent accuracy audits, requires privacy commissioner review before any live deployment, and sets hard limits on data retention and sharing would give Albertans something they currently do not have: confidence that the technology serving public safety is itself subject to public accountability.
Should Alberta pass biometric policing legislation before any facial recognition program moves beyond a pilot phase?
