AI OK but Biometrics Not

A small article in the news section of March’s ID & Secure Document News noted that at the same time as European Union lawmakers are on track to ban the use of remote biometric surveillance for general law enforcement purposes, France had approved the deployment of artificial intelligence (AI) to monitor public spaces for suspicious behaviour during the 2024 Paris Olympics.

Following that, the Constitutional Council, the highest constitutional authority in France, scrutinised the provisions of the law and approved the use of algorithmic processing of video feeds during the Olympics but restricted the system from being used to collect biometric data.

The provisions that the government sought to improve safety at the Olympics are, in effect, an extension of a previous article that allowed for ‘images collected by means of a video protection system or cameras installed on aircraft to be subject to algorithmic processing in order to detect and report certain events.’

The provisions contested by the Council were more specifically to do with images collected in places hosting certain events and in their surroundings that may be subject to algorithmic processing aimed at detecting and reporting acts of terrorism or other serious threats to public safety.

These provisions were criticised for disregarding the freedom to come and go, the right to demonstrate, freedom of opinion and the right to respect for private life. In support of these grievances, the Council considered that the use of algorithmic processing was not surrounded by sufficient guarantees. In particular, it claimed that the scope of these provisions, which is not limited to events linked to the Olympic and Paralympic Games, was too broad and that the detection of certain events would necessarily lead to the processing of biometric data even though the law forbids it.

The Council decided that the algorithmic processing of the images thus collected may only be implemented in order to ensure the safety of sporting, recreational or cultural events which, by the extent of their attendance or by their circumstances, are particularly exposed to the risk of acts of terrorism or serious threats to personal safety.

However, a limit has been placed on the use of processing of images ‘so that the algorithmic processing does not implement any facial recognition technique, does not use any biometric identification system and does not have recourse to biometric data, that is to say relating to physical, physiological or behavioural characteristics of a natural person which allow or confirm their unique identification. Furthermore, the processing may not carry out any comparison, interconnection or automated linking with other processing of personal data.’ 

Interestingly, there was also a concern expressed regarding bias: ‘throughout their operation and in particular in the case where they are based on learning, the algorithmic processing used must make it possible to verify the objectivity of the criteria adopted and the nature of the data processed as well as include human control measures and a risk management system to prevent and correct the occurrence of any biases or misuse.’ 

The ruling of the Council is of interest to ID & secure document professionals as it is indicative of the trend in public opinion that is suspicious of facial identification for surveillance purposes (unless a clear benefit to public order and safety can be demonstrated) and how this fear of surveillance might be conflated with a fear of identity authentication.