First, promised juror, it was only for the police. Then hundreds of companies were caught in the bag by Buzzfeed. Finally, the New York Times revealed on Thursday, March 5, that a club of “happy few” - wealthy investors, friends - also had access to the facial recognition application of one of the most successful start-ups. controversies of the moment: Clearview AI.

In a few seconds and a stolen photo, the application enabled a rich billionaire to find the identity of a stranger who was dining with his daughter. Or, the police identify a sexual predator, absent from official databases but whose facial reflection appeared in a mirror on a photo taken by someone in a gym. These are just two examples of the applications of this controversial service.

Nightmare for privacy, boon for the police

Clearview AI is the story of a New York start-up that has done everything for years to remain as discreet as possible, aware that its facial recognition technology was pushing the limits of morally acceptable very far.

An investigation published by the New York Times at the end of January put an end to this anonymity. Revelations that have earned the start-up a flock of criticism from the media and major Internet giants, such as Google, Twitter or Facebook.

The software engineer me: Wow this is amazing tech!

The American citizen me: JFC this is scary tech. https://t.co/8dLjAXzJHE

- Christopher Bouzy (@cbouzy) January 18, 2020

The Reasons Of The Wrath ? About three billion free access photos that Clearview AI has gleaned from its creation in 2016 on most social networks and millions of websites to build an unparalleled database. A homemade algorithm then takes care of quickly and efficiently comparing a snapshot taken with a smartphone with the gigantic war chest of Clearview AI.

The tool seduced the American police. Starting with the Indiana police who became the first client of the start-up in February 2019. For the agents, there was no photo: until then, they only had access to software much more limited facial recognition, which allowed for cross-checking with image banks comprising “only” tens of millions of photos of individuals who had already dealt with the police. In addition, the face in the photo to be compared must have been perfectly clear while Clearview software finds a person even if part of the face is hidden. A feat for facial recognition software.

The results not only indicate the name, but also provide all the information associated with the photo on the Internet, such as professional occupation if the snapshot is associated with a LinkedIn profile, for example.

A privacy nightmare, but a terribly effective weapon that has been used by about 600 local police in the United States, the New York Times has learned. The app has helped to find fraudsters, shoplifters, and pedophiles. It was even used to put a name on an unidentified corpse.

For Hoan Ton-That, the boss of Clearview AI, these police successes justify his controversial innovation. After the publication of the first New York Times investigation, he toured the American media to reiterate that his app was only intended for the police. And only in the United States and Canada. The start-up has also put online a “code of conduct” explaining that the purpose of the application was only to help “law enforcement and certain security professionals” to “identify victims and the criminals ”.

From Saudi Arabia to Ashton Kutcher

A claim undermined by new revelations published by Buzzfeed in late February and the New York Times thereafter. Walmart, Best Buy, Bank of America, the NBA, or even Las Vegas casino chains appear in the Clairview AI user file, consulted by Buzzfeed. In all, nearly 200 companies have done hundreds of thousands of searches using this facial recognition software, as part of the start-up's free trial program. There is nothing to indicate, however, that they subsequently became paying customers.

To attract investors, the start-up has also abandoned its "code of conduct" by allowing wealthy businessmen to have fun with its software, the New York Times learned. “I made it a real hobby. I ask permission from people I meet. It's like doing a magic trick! ”, Tells the American daily newspaper Nicholas Cassimatis, an expert in artificial intelligence. Actor and investor Ashton Kutcher also seems to have had the opportunity to use it. During an interview in September 2019, he spoke about this “application that I have on my smartphone and which identifies anyone. It's terrifying".

Buzzfeed adds that Clearview AI is starting to feel too cramped within North American borders. The software has been promoted to authorities in more than 20 countries. The American start-up has approached both democracies, such as Australia and France, as well as more authoritarian states, such as Saudi Arabia or the United Arab Emirates. Again, the documents consulted by Buzzfeed do not indicate whether some of these countries have become paying customers.

Enough to make Google, Facebook and Twitter even more nervous that photos posted on their service could be used by authoritarian regimes to hunt dissidents. The major platforms have all asked Clearview AI to stop looting their images and erase those already recovered. Hoan Ton-Than opposed it, arguing that access to public data is a matter of freedom of expression and information.

Clearview AI may eventually give in, ending its service. But “there is no monopoly on algorithms, said Al Gidari, professor of personal liberties at Stanford University, interviewed by the New York Times. We may find it very disturbing, others will do it [if Clairview stops, note]. Without a very strict facial recognition rule, we're all screwed up. ”

In the meantime, Clearview AI seems to have ideas to make its software even more efficient ... and invasive for privacy. In its facial recognition program, consulted by the New York Times, there is already a code that would allow the software to be adapted to augmented reality glasses. What identify any person crossed in the street? Hoan Ton-That admitted to having thought about such an option, before giving up. Promised sworn.

The France 24 week summary invites you to come back to the news that marked the week

I subscribe

Download the app

google-play-badge_FR