For three months, the Française des jeux has been testing a new way of verifying the age of its customers: with facial recognition software.
After Instagram, Facebook Dating and Yubo, it’s the turn of the Française des jeux to test facial recognition to verify the age of its customers. The company, which publishes gambling and games of chance, announced on Wednesday April 5 that it had launched a test campaign in several tobacconists in France, reported l’Usine Digitale.
For the past three months, tobacconist partners have installed terminals in their shops, equipped with a camera and facial recognition software. The latter estimates the age of customers, who must not be under 18, thanks to the analysis of biometric data – and thanks to artificial intelligence.
Results provided by facial recognition software
For the age analysis to be done, ” the player only has to position himself in front of the camera for a photo to be taken and analyzed “, details the Digital Factory. Concretely, the software is based on the biometric data of the faces to make its analysis, and its AI, which resulted in ” thousands of individuals of different age, sex and skin colore”, according to 01Net, allows conclusions to be drawn.
The result of the analysis, which is given after only one second, is very precise, reassures France of the games: the margin of error is 1.6 years. L’Usine Digitale states that “ the system was set to alert the tobacconist from the age of 20. Below, the person must present his identity card to avoid any errors.
The FDJ uses software developed by Yoti, an English start-up, to make its estimates. Yoti, which was created in 2017, is already popular with big names in tech: it is their software that has been chosen by Instagram, Facebook Dating and Yubo to verify the age of their users. It is also Yoti who has been chosen by PornHub to verify the age of users posting videos on the site, and its software is used in England in automatic supermarket checkouts for the purchase of alcohol.
Facial recognition still makes mistakes
We are therefore dealing with an expert in the sector. However, there are still a few problems: even if the company indicates that the average error is 1.6 years, the difference can be large. Research carried out by Yoti in May 2022 had shown that the margins of error were greater for women with dark skin: the average error goes up in some cases to 4.79 years.
What about data processing? La Française des jeux and Yoti ensure that neither the photos taken nor the data collected have been kept. The English startup also insists that its software does not identify users, only verify their age. ” We cannot sell your data to third parties “, reassures the site of Yoti. ” Once we have done the analysis, we can no longer access user data “.
Above all, there remains the question of the success of the tests. The FDJ has not taken stock of this 3-month test phase: there are no published figures attesting to a faster or error-free process, or even if the test terminals attracted a lot of volunteers. The company, questioned by AFP, just specified that ” majority image analysis and certification solutions with Yoti » were still only « prototypes “.
Subscribe to Numerama on Google News to not miss any news!