On November 15, 2019, the French Supervisory Authority (“CNIL”) published guidance on the use of facial recognition. The guidance is primarily directed at public authorities in France that want to experiment with facial recognition.
The guidance warns that this technology risks leading to biased results because the algorithms used are not 100% reliable and the rate of false-positives may vary depending on gender and skin color. It sets out three general requirements for deploying facial recognition on an experimental basis.
First, facial recognition can only be used if there is an established need to implement an authentication mechanism that ensures a high level of reliability, and there are no other less intrusive means that would be appropriate. In this regard, the guidance lists examples in which the CNIL has found the use of facial recognition to be lawful. These include the use of facial recognition to access public services online (e.g., tax account, health insurance account, vehicle registration, driving license, passport or ID card) and automated identity verification systems used at borders and travel hubs, such as airports. In contrast, the CNIL decided against the use of facial recognition in schools for security and access purposes.
Second, the experimental use of facial recognition must respect the rights of individuals, which translates into a series of requirements. For example, an organization must obtain consent from individuals for each device used to carry out facial recognition, in particular when used on an experimental basis. Individuals must also be given control over their data and receive clear, comprehensive and sufficiently accessible information. The guidance indicates that the use of facial recognition on an experimental basis should not have “the ethical purpose or effect of accustoming people to intrusive surveillance techniques, with the more or less explicit aim of preparing the ground for further deployment”. This will only be permitted at a later stage for devices that are recognized as “perfectly legitimate and legal”.
Third, the use of facial recognition on an experimental basis must have a precise timeline and be based on a rigorous methodology setting out the objectives pursued and the criteria for success.