AI powered age verification systems are now being tested at a few UK supermarkets to verify the ages of people attempting to buy alcohol at self-checkout.
UK supermarkets have begun testing an automated age verification system at self-checkouts when buying alcohol. According to this BBC report, this system will utilize cameras that can estimate a customer’s age. The system is being installed in some supermarkets and will require any customer who is estimated to be under 25, to present ID to a staff member in order to check out. The system uses a camera which will guess their age using algorithms trained on a database of anonymousfaces. Customers must consent to the use of this system.
Producers of this system maintain that it is not facial recognition and that images are not retained.
The system is being produced with the intention of speeding up the self-checkout process by eliminating the need to wait for a member of staff to verify the ages of customers using self-checkout. The producer, a company called Yoti, has stressed that this system is not facial recognition and that images taken by the cameras are not saved. This is intended to protect customers’ identities. Unlike facial recognition, the system does not match faces to individuals’ faces from a database, but rather compares the face to the anonymous faces in the database in order to guess a customer’s age. According to Robin Tombs, chief executive of Yoti, “Our age-verification solutions are helping retailers like Asda meet the requirements of regulators worldwide and keep pace with consumer demands for fast and convenient services, while preserving people’s privacy.”
The producers of this AI powered age verification systems claim that scanned images are never stored, nor are they ever shown to anyone.
This system has been tested on over 125,000 faces between the ages of six and 16. On average the system was able to guess the age of participants to within 1.5 – 2.2 years among 16-20 year olds. While there have been several concerns about privacy with regard to the use of facial recognition in public spaces, the producers of this age verification system maintain that this system is not the same as facial recognition, and that their personal data is not processed, and images are not saved. The company’s website states that the facial age estimation system can be embedded into an app, website or POS terminal. The AI powered technology can estimate an individual’s age within seconds without the need to provide identification, and without the need to store the image. It goes on to note that the image will never be seen by anyone.
The system might not be free from potential compliance issues though. According to Dr Bostjan Makarovic, Aphaia’s Managing Partner, the system still processes one’s facial features and perceived age, then bases an automated decision on one’s right to buy a certain item upon such profiling. The implementation of one’s right to obtain human intervention will therefore be crucial.
“These systems may also raise discrimination concerns if the data used to train them was not high-quality data, therefore an AI ethics assessment may also be required to ensure full compliance” points out Cristina Contero Almagro, partner in Aphaia.
