Weak Privacy, Weak Procurement: The State of Facial Recognition in Canada

Executive Summary

Following the Clearview AI scandal, this report examines the state of facial recognition technology (FRT) in Canada, arguing that technology companies and law enforcement are taking advantage of weak privacy and procurement regimes to increase capital and power under the pretext of public safety. Over 30 police departments in Canada used free trials of Clearview AI's software without public knowledge, and by the time the scandal broke, Canadian police had already run 3,400 searches across 150 accounts. By February 2021, four privacy commissioners across Canada ruled that Clearview AI continues to engage in "mass surveillance" by scraping photos of people residing in Canada and refusing to delete images it had collected.

Drawing on the analytic framework of sociotechnical security, the authors identify two primary vulnerabilities in Canada's legal regimes: the omission of biometric information as a weakness in Canadian privacy law that prizes organizational efficiency over the protection of dignity, and the ease with which companies like Clearview AI can exploit weak transparency requirements when public bodies enter into contracts with private companies. The report also surveys evidence of FRT's discriminatory harms, including research finding error rates of up to 34.7 percent for darker-skinned females compared to a maximum of 0.8 percent for lighter-skinned males, and findings that 189 facial recognition algorithms were 10 to 100 times more likely to inaccurately identify Black and East Asian people.

The authors conclude that these gaps in privacy and procurement law leave Canada's marginalized communities most exposed to the harms of unaccountable FRT use, and recommend mandatory privacy impact assessments and algorithmic impact assessments with consequences for non-compliance, as well as broader application of legislation such as the Directive on Automated Decision-Making across Canadian jurisdictions.