How a secretive startup’s facial recognition technology became the embodiment of our dystopian fears
In November 2019, while working as a reporter at The New York Times, Kashmir Hill uncovered a story that would expose one of the most controversial developments in surveillance technology.
As reported here, journalist Kashmir Hill recalls the rise of Clearview AI. This facial recognition technology company gained widespread attention with its artificial intelligence software that claimed to be able to identify almost anyone with a single picture of their face, in this excerpt from “Your Face Belongs to Us” (Simon & Schuster, 2023).
Clearview AI, an enigmatic startup, promised to be able to identify almost anyone from a picture of their face.
According to some rumors, Clearview had scraped billions of photos from the public web, including social media sites such as Facebook, Instagram, and LinkedIn, to create a revolutionary app.
A random person’s name and other personal information about their life may be revealed if you show Clearview a picture of them taken on the street. It would then spit out all the websites where it had seen their face. While attempting to conceal its existence, the company sold this superpower to police departments nationwide.
Until recently, most people thought that automated facial recognition was a dystopic technology only found in science fiction books or films like “Minority Report.” To make it a reality, engineers first tried programming an early computer in the 1960s to match a person’s portrait to a wider database of faces. Police started experimenting with it in the early 2000s to look up the faces of unidentified criminal suspects in mug shot databases. But for the most part, the technology had fallen short. Even cutting-edge algorithms had trouble matching a mug image to a grainy ATM surveillance still, and its performance differed depending on age, gender, and color.
Claiming to be unique, Clearview boasted a “98.6% accuracy rate” and a vast photo collection that was unmatched by anything the police had previously employed.
In 1890, a Harvard Law Review article famously defined privacy—a term that is notoriously difficult to define—as “the right to be let alone.” Samuel D. Warren, Jr. and Louis D. Brandeis, the two lawyers who wrote the article, argued that the right to privacy should be legally safeguarded in addition to the previously established rights to life, liberty, and private property. They were influenced by then-novel technology, such as the Eastman Kodak film camera, which was introduced in 1888 and allowed one to shoot “instant” pictures of everyday life outside of a studio.
“Instantaneous photographs and newspaper enterprise have invaded the sacred precincts of private and domestic life,” wrote Warren and Brandeis, “and numerous mechanical devices threaten to make good the prediction that ‘what is whispered in the closet shall be proclaimed from the house-tops.'”
Louis Brandeis later joined the Supreme Court, and this essay is one of the most popular legal essays ever published. However, privacy never received the level of protection that Brandeis and Warren claimed it deserved. There is still no comprehensive law that ensures Americans have control over what is written about them, what is photographed of them, or what is done with their personal information more than a century later. In the meantime, companies in the US and other nations with weak privacy regulations are developing increasingly powerful and intrusive technology.
Examples of facial recognition include digital billboards from Microsoft and Intel that use cameras to detect age and gender and display relevant advertisements to onlookers, Facebook that automatically tags friends in photos, and Apple and Google that allow users to unlock their phones by looking at them.
In a matter of seconds, a stranger at a bar may take your picture and determine your friends’ identities and residences. It might be used to track down women who entered Planned Parenthood facilities or anti-government demonstrators. It would be used as a tool for intimidation and harassment. The third rail of the technology was accurate facial recognition for hundreds of millions or even billions of people. Now Clearview has made it.
We tend to think of computers as having nearly magical abilities, capable of solving any problem, and, with enough data, eventually outperforming people. Therefore, companies that want to produce something amazing but are not quite there yet can deceive investors, customers, and the general public with ludicrous statements and certain digital tricks.
However, Paul Clement, a prominent lawyer for Clearview and former US solicitor general under President George W. Bush, said in one private legal memo that he tested the system with lawyers from his company and found that it provides fast and accurate search results.
According to Clement, the tool is currently being used by over 200 law enforcement agencies, and he has concluded that when using Clearview for its intended purpose, they do not violate the federal Constitution or any existing state biometric and privacy laws. In addition to the fact that hundreds of police departments were secretly using this technology, the company employed a high-profile lawyer to convince officers that their actions were not illegal.
For decades, worries about facial recognition have been building. And now, at last, the unidentified monster had taken the shape of a small company with enigmatic founders and an enormous database. Furthermore, none of the millions of individuals that comprised that database had provided their approval. Although Clearview AI embodies our darkest anxieties, it also provides the chance to finally face them head-on.
The 2019 launch of Clearview AI signaled a turning point in the continuous conflict between privacy and technical progress. Clearview AI’s unparalleled database and precision brought these gloomy worries to stark reality, even though facial recognition had long been confined to science fiction and a few law enforcement uses. As the company carries on and grows, it now acts as a warning and a vital impetus for tackling the pressing need for all-encompassing privacy laws in the digital era.
In addition to exposing a controversial company, the legal document that arrived in Hill’s inbox revealed a future that privacy advocates had long dreaded and cautioned against. The question of whether such tools will exist is no longer relevant when we consider the ramifications of this technology; rather, it is how society will decide to control and limit them. We are reminded that the “right to be let alone” is still as important—and possibly as vulnerable—as it was more than a century ago by Warren and Brandeis’s 1890 warning against invasions of privacy.