Toronto police using facial recognition as Canadian government ponders rules
Toronto police used facial recognition technology to attempt to identify people in 2,591 searches since March of last year, according to a report by Chief Mark Saunders which revealed the force’s use of the technology, the Toronto Star reports.
A report submitted to the Toronto police services board shows that images from public and private cameras are matched against an internal database of 1.5 million mugshots, and that the system’s use so far has cost CAD $451,718 (just over US$335,000). According to Saunders, the system was purchased to help police more quickly and efficiently identify suspects, including violent offenders. A provincial grant for police modernization funded the purchase.
New Democratic Party Member of Parliament Charlie Angus told the Star that no legislative oversight is in place, and therefore “we need to hit the pause button.”
Angus is also part of the House of Commons Standing Committee on Access to Information, Privacy and Ethics, which is currently studying artificial intelligence ethics. Angus told Mobile Syrup that his office is considering legislative changes to restrict facial recognition.
Facial recognition is limited in Canada by the Personal Information Protection and Electronic Documents Act (PIPEDA), which requires consent for the collection, use, or transmission of personal information, but does not specifically deal with AI or facial recognition.
“I think we really need to look at putting limits on facial recognition technology and lay the ground rules before it gets widely implemented,” Angus says.
Toronto police’s implementation generated matches for roughly 60 percent of 1,516 searches between March and December of 2018, about 80 percent of which led to the identification of criminal offenders, according to the report to the police services board. The report also states that facial recognition is not used as a sole basis for arrests, unlike fingerprint identification, but rather to produce potential candidates for further investigation. Information provided by facial recognition helped to solve four homicides, multiple sexual assaults, armed robberies, and gang-related crimes.
The Star asked Toronto Police Services about false positive rates overall and for different ethnic groups, and was told the technology is not used to make a positive identification. A representative of the police also said the force has no plans to extend its database beyond the existing mug-shot collection, it does not use real-time facial recognition, and does not have legal authorization to do so. Only six FBI-trained officers have access to the system, while body camera images can be used in the case when a criminal offence is captured on camera, court permission is still required.
Saunders’ report notes that the police began a year-long pilot project in September, 2014, and conducted a Privacy Impact Assessment in 2017.
“The fact that there has been very little — virtually no — public conversation about the fact that this is happening, despite the fact that they’ve been looking into it for at least the past five years … raises questions for me,” the Canadian Civil Liberties Association’s Privacy, Technology and Surveillance Project Director Brenda McPhail told the Star.
Calgary is the only other Canadian city where police are known to use facial recognition technology.
No federal legislation is imminent, but the house committee is attempting to set up the next sitting of parliament to move the issue forward, according to Angus.
Liberal MP and committee co-chair Nathaniel Erskine-Smith told Mobile Syrup that there are benefits to some AI technologies, but acknowledged that negative outcomes could necessitate a ban.
“Where it is not mitigated and in the case of San Francisco, if there is clear evidence that employing the algorithm leads to racial profiling then governments, be it local or national, should prohibit the use of that technology,” Erskine-Smith says.
Canada’s government recently set up an Advisory Council on Artificial Intelligence.