Skip to main content

Toronto Police Service denied it was using Clearview AI, even as some cops used it for four months, unbeknownst to the senior brass.Amr Alfiky/The New York Times News Service

The 18th-century philosopher Jeremy Bentham came up with the idea of a building he called the panopticon. It was a prison laid out in such a way that one guard could view all of the jailed. Inmates would never know, from moment to moment, if they were being watched. The threat of constant surveillance aimed to cow prisoners into pliancy.

The potential for mass surveillance – a panopticon far beyond Bentham’s imagination – is a 21st-century reality, after advances in online media, artificial intelligence and facial recognition. China is able to exercise nearly invisible yet authoritarian control over 1.4 billion people with omnipresent mass surveillance. There is one security camera for every six Chinese, and widespread use of facial recognition for everything from policing to people accessing their own homes.

In Canada, that camera at the intersection, or behind a convenience store counter, is not an all-seeing eye, constantly scanning faces. But the lure of such technology is strong enough that four dozen organizations in Canada, most of them police forces, have used facial recognition software from a small American company called Clearview AI, which has compiled a huge universe of photo data.

Last week, after a year-long investigation by federal and provincial privacy commissioners into the use of Clearview, Ottawa’s Privacy Commissioner Daniel Therrien was blunt: “What Clearview does is mass surveillance, and it is illegal.”

The upstart company was unknown until a New York Times story in early 2020. Its co-founder had dabbled in iPhone apps. One failed effort was “Trump Hair,” which allowed users to get a laugh by adding The Donald’s distinctive mane to their own photos. Then came the idea for Clearview. The company downloaded more than three billion images of faces from the internet, marked 512 data points on each to make the database searchable, and pitched it to police.

Canadian police were eager and it often was individual officers signing on. In one example, the Toronto Police Service denied it was using Clearview, even as some cops used it for four months, unbeknownst to the senior brass. The privacy investigation revealed one unnamed police force conducted thousands of searches. Clearview exited the Canadian market last summer, during the probe.

The company insists the images it gathered online were in the public realm. The privacy commissioners disagree. An image may seem to be sitting around in “public” on the internet, but no reasonable person would argue that provides consent to copy it, put it in a database, and sell it to police. When a person posts a photo on Facebook, they don’t think they are agreeing to put it, and themselves, in an infinite digital police lineup.

Personal privacy is a pillar of a liberal society. Your life is for the most part not somebody else’s business, or the government’s. Just because someone puts your picture online, or you walk down a street, doesn’t necessarily make your face public property. Biometric information is sensitive data, and the privacy commissioners note that “facial recognition data is particularly sensitive.”

And it’s not just about policing. Last fall, Canadian mall owner Cadillac Fairview was found to have violated privacy laws after it used covert cameras and facial recognition to analyze shoppers. A few places, such as San Francisco, have banned facial recognition. Beyond privacy, there are concerns about bias and accuracy.

The fact that for the first time in human history, technology allows the creation of databases of trillions of images, e-mails, phone calls and location-tracking data, all searchable, merely shows what is technically possible. It doesn’t say what is legal, or ethical or desirable.

New technologies cannot be uninvented. But some old principles of law can help us think about appropriate limits that should be placed on their use. Canadians have long enjoyed the right to privacy in their homes and security of their personal effects. Police cannot ask to look inside the purse of every person on the street, or barge into random homes, searching for signs of crime. A warrant, based on evidence, is necessary. New tech must be likewise limited.

The federal and provincial privacy commissioners this spring will present initial guidelines for police. It’s a start. The frightening power of facial recognition has rapidly morphed from science fiction to reality. Its dystopian potential is obvious. Just look at China.

Keep your Opinions sharp and informed. Get the Opinion newsletter. Sign up today.