The National Assembly has just passed a law authorizing the use of artificial intelligence for augmented video surveillance during the 2024 Olympic Games. A decision that is debating and which worries the CNIL…
Video surveillance has become part of our daily lives, and it’s not about to stop. While video surveillance systems initially relied on analog devices (cameras, video recorders, etc.) and human operators, they are increasingly using automated digital technologies, which are more efficient but also more worrying, at a time when the artificial intelligence (AI) is going mainstream. There National Commission for Computing and Liberties (the famous CNIL) had already issued a warning in July 2022 against the deployment of so-called “augmented” cameras, which are increasingly used to monitor public space, pointing out the risk to the privacy of citizens (see our article). A risk which is likely to increase with the law relating to the 2024 Olympics, voted on Thursday March 23 by the National Assembly, which aims to create new mechanisms to regulate the sporting event.
Indeed, this law relates to many points, such as the arrangement of the opening of stores on Sundays, the extension of the powers of the Paris police chief to certain departments, the use of body scanners with consent, the carrying out genetic doping controls, the strengthening of sanctions for violence in stadiums… and the use of artificial intelligence (AI) in the field of video surveillance. A use that greatly worries the CNIL, the left and the associations, but which was nevertheless authorized with 59 votes for (presidential majority – LR – RN) against 14 votes against (Nupes), as reported The world. This law provides that the securing “sporting, recreational or cultural events” scale can use algorithms. It will be tested before and during the 2024 Olympic Games. The list of “events” to be detected must still be fixed by decree, after consulting the CNIL. During the debates, Interior Minister Gérald Darmanin cited in particular “a fire outbreak, population bottlenecks, an abandoned package or bag”but “not the hoodies”.
Video surveillance by AI: an experiment set to last
Article 7 sows discord by authorizing the installation of experimental automated video surveillance devices, “which will use artificial intelligence (AI) algorithms capable of detecting situations dangerous to the safety of people, such as crowd movements”, but also abandoned packages, “abnormalities” and suspicious movements. These “intelligent” cameras – also called augmented cameras – analyze the captured images in real time in order to“ensure the security of sporting, recreational or cultural events (…) which, by their scale or their circumstances, are particularly exposed to the risk of acts of terrorism” . We therefore understand that this system will also concern the Rugby World Cup in May 2023 and, above all, will be required to remain permanently.
“AT exceptional situation, exceptional means”, said Gérald Darmanin. The Government justifies the use of AI by the too many cameras installed, because “live viewing of all the images captured by video protection cameras is materially impossible” and artificial intelligence would make them “save valuable time”as reported 20 minutes. In reality, it would also be for economic reasons, with the Olympics serving as a true global showcase for camera and software companies. Also, even if the experiment is in theory supposed to end on June 30, 2025, it seems set to last. THE Board of state pointed out, moreover, in an opinion published in December, that 11 of the 19 measures of the text presented “a permanent character” and were “designed to apply even outside the period” of the Olympic and Paralympic Games.
Biometric cameras: the risks of political drift
The CNIL shares the opinion of the Council of State and considers that “the deployment, even experimental, of these devices constitutes a turning point which will contribute to defining the general role which will be attributed to these technologies, and more generally to artificial intelligence“. For their part, the associations are on a war footing. “The government is using the Olympics as a pretext to pass measures aimed at accelerating surveillance of the population,” was already alarmed The Squaring of the net in January. “The bill proposes to experiment with automated video surveillance even though no public evaluation of current video surveillance devices exists (…) nor any scientifically demonstrated utility”, advances the association, citing an rcontribution of the Court of Auditors, according to which “no overall correlation has been found between the existence of video protection devices and the level of crime committed on the public highway, or even the rates of elucidation”.
In a press release published in December, the CNIL asked for guarantees in order to prevent things from getting out of hand and France from ending up resembling the Chinese model and its biometric identification, which notably makes it possible to identify individuals directly in the street – very practical for targeting minorities, the marginalized, the whistleblowers and any other opponent of the regime. Facial recognition, which matches a human face to a digital image using scans and CCTV cameras – and which has already been adopted by eleven countries in the European Union, particularly in a judicial setting – is particularly feared. The digital policeman therefore recommended “the absence of biometric data processing” and of “reconciliation with other files“, which the Senate had accepted.
Gérald Darmanin heavily insisted that it would be neither facial recognition nor biometric data. For him, it is “only” to identify predetermined situations so that the operator decides if they deserve an intervention. However, he ran into skepticism from some MPs. “The Defender of Rights explicitly recalled that the detection of so-called abnormal behavior was based on biometric data“, underlined Lisa Belluco (Europe Ecology-The Greens). “The algorithm will make it possible to recognize people, without necessarily identifying them. Recognizing means providing a sufficiently detailed description to allow agents in the field to locate a person”. Europe also does not see this decision very favorably. On March 17, some forty MEPs expressed their disagreement with Article 7, arguing that by adopting this provision, without waiting for the results of the debates at European level, France “risk of coming into conflict with European law” on artificial intelligence, which should arrive by the end of the year. And according to a 2021 report, the European Parliament would rather for “the permanent ban on the use of automated analysis (…), human characteristics (…), and other biometric and behavioral signals”. This could be problematic…