La Française des jeux (FDJ) tested for three months an artificial intelligence, developed by the British start-up Yoti, capable of evaluating the age of consumers, the latter announced on Wednesday April 5. This technology is based on the analysis of a person’s facial features. At the tobacconists, young customers had the choice between presenting their identity document or letting a computer scan their face to prove that they were of legal age. The goal? Protect the youngest by prohibiting them from gambling.
“We train the machine by showing it lots of photos of people whose ages we already know,” explains Iain Corby, executive director of the World Association of Age Assessment Solution Providers. Once operational, it is then able to estimate a person’s age with a margin of error of one and a half years. “It’s pretty reliable, especially with young people,” he says. To limit errors, the FDJ has placed its limit at 20 years. If the terminal considers that the person is younger, an identity document is requested.
A technology that is gaining ground
Yoti is now a member of the Laboratory for Online Child Protection, launched in November 2022 by the Élysée. During the International Cybersecurity Forum, organized in Lille on April 5, 6 and 7, it affirmed its ambition to see its technology used on pornographic sites. This has been tested on Instagram in the European Union since November to verify that users are over 13 years old, registration on the application being prohibited below this age.
For Cédric Vasseur, specialist in new technologies related to artificial intelligence and robotics, this tool can have a reassuring aspect for the population. “Showing your ID is like revealing who you are, what you’re doing, you feel like you’re being tracked,” he analyzes. For its part, the technology is content to send an age to the servers, the photo not being kept. Personal data is thus protected.
Potential abuses
But if the speaker is convinced of the usefulness of this technology, he remains aware of its limits, particularly in terms of privacy. “You have to pay attention to the systems usedhe warns. There may be fake ones, which store the photos or even divert them. » To avoid this, he calls for vigilance and recommends the use of public and well-documented algorithms.
As for the reliability of such technology, Cédric Vasseur is on the reserve. “There is no AI capable of estimating a person’s age to the day. A year, and even six months apart, is already huge. » In the event that such technology is used to control the sale of products prohibited to minors, the responsibility of the cashier would then be engaged. It is up to him to ask for an identity document if there is any doubt about the age of the person to avoid the illegal purchase of a product. And on the Internet, the speaker fears that teenagers will find tricks to access content that is forbidden to them.
For the moment, this technology is only used in the form of a test in France. Cédric Vasseur specifies, however, that it is already used at European borders to verify that undocumented people arriving on the continent are indeed of the age they claim to be.