June 6, 2026
Algorithmic Video Surveillance during the Paris Olympic Games
The Conseil d’État (Council of State), in an article on security, states: “Finding the right balance between security and the preservation…
42 Artificial Intelligence
6 min read
The Conseil d'État (Council of State), in an article on security, states: "Finding the right balance between security and the preservation of freedoms is essential for the rule of law. Through its decisions and opinions, the Conseil d'État ensures that freedoms are respected and that any potential infringements upon them are always justified and proportionate." In other words, the highest administrative court considers that a compromise must be made between freedom and security.
This position is, in reality, not surprising. It was theorized by philosophers such as Hobbes and Rousseau, and serves as the ideological foundation for most democracies. Infringements on freedoms, which can be likened to a form of violence, are only legitimate as long as they serve to preserve other freedoms.
The Principle of Proportionality
The principle of proportionality requires judges — whether French or European — to verify that an infringement on a fundamental right is not disproportionate.
First, the judge must verify whether the measure pursues a legitimate aim, then whether it effectively achieves that aim, and finally, whether another less restrictive but equally effective measure could have been taken instead.
It also helps resolve conflicts between opposing fundamental rights, such as freedom of expression versus the right to privacy. This is done on a case-by-case basis by weighing the interests involved, either to reconcile them or to allow one to prevail over the other depending on the specific circumstances of the case.
Definition from the website of Dalloz, a publishing house specialized in law.
Technological upheavals, such as the advent of artificial intelligence, raise ethical questions because they expand the scope of possible security actions. Furthermore, the principle of proportionality used to justify their intrusion is eminently subjective. A particularly polarizing example of these new paradigms is Algorithmic Video Surveillance (AVS) in public spaces for policing purposes.
The promise of these new technologies is the ability to identify high-risk situations more frequently and rapidly. In return, we consent to the installation of image capture and processing systems in certain public spaces. To determine whether the means implemented justify the infringements on freedom, let us examine the experimentation that took place during the 2024 Paris Olympic Period (JOP).
The Paris Olympic Games: A Particularly High-Risk Event
Faced with security fears linked to the terrorist threat, a shortage of personnel, and the sheer scale of the event, the use of Algorithmic Video Surveillance solutions was authorized a few months before and during the Paris Olympics. The first operational use of these cameras took place in April 2024. However, this deployment occurred within a strict framework that is important to clarify to avoid erroneous analyses.
First, identification via biometric data was completely ruled out. Surveillance solution providers were therefore prohibited from determining the identity of the people filmed. This is why the CNIL (French Data Protection Authority) chose to nickname these systems "augmented" cameras. This infrastructure was exclusively designed to detect:
- The presence of abandoned objects
- The presence or use of weapons
- A person or vehicle traveling against the traffic flow
- The crossing of a boundary, or the presence of a person or vehicle in a prohibited or sensitive area
- A person on the ground following a fall
- Crowd surges
- Excessive crowd density
- The outbreak of a fire
Second, the deployment of these solutions had clear temporal and spatial limits. Authorizations to use AVS were granted for sporting, recreational, or cultural events only when attendance or circumstances suggested a significant risk. The CNIL acknowledged that these systems could clash with the GDPR and the exercise of many individual freedoms (freedom of movement, expression, assembly, the right to protest, freedom of worship, etc.) if deployed in certain public spaces without clear limits. Furthermore, these video systems were only meant for real-time detection. They were not allowed to use the collected data to improve their models or to issue retroactive alerts.
Third, to evaluate the relevance of these solutions, a briefing report by the Law Committee was published on February 19, 2025. It aimed to determine the operational impact of these systems to assess whether their use aligned with the principle of proportionality. Since the deployment of AVS was only made possible within an experimental framework (Article 10 of the Law of May 19, 2023, relating to the 2024 Olympic and Paralympic Games), this evaluation was mandatory. It is also important to note that it was the result of cross-party collaboration, notably featuring a Republican and a Socialist representative as rapporteurs.
According to a report by the National Assembly, the Olympic Games took place without any major incidents, despite numerous fears, particularly regarding the outdoor opening ceremony. However, this does not mean that AVS was the cause of this success. It is vital to keep in mind that the entire security apparatus was reinforced during the Games. For instance, approximately 26,000 private security agents were trained and deployed for the occasion.
Algorithmic Video Surveillance: A Success?
The Law Committee report mentioned above paints a mixed picture of the AVS deployment.
From a formal standpoint, these systems mostly operated within the legal framework imposed upon them. Data processing took place within the boundaries set by law. Public communication regarding its use was identified as the major area for improvement; it was sometimes deemed belated and should have been more educational.
However, their effectiveness could not be confirmed for several reasons, starting with technical constraints. The timeline provided by the experiment did not prove particularly suited to these technologies, which require a significant calibration phase before they can be effective. This step is crucial for reducing false alerts. Since the law permitting the experimentation was adopted at the end of 2023, it left very little time to calibrate the devices.
The sample size of observations was also limited, preventing definitive conclusions. The total number of private AVS companies was restricted to three suppliers selected for the trial, and only one was used on a large scale. Consequently, it becomes difficult to determine the ultimate effectiveness or ineffectiveness of these systems.
"The calibration phase consists of verifying the quality of the initial programming, both for all equipped cameras combined and on a camera-by-camera basis. At this stage, the solution is activated under real conditions but without any operational exploitation. This step allows the programming to be adjusted if necessary to limit the number of false alarms and to fine-tune detection thresholds based on operational needs."
— Senate Report
From an operational perspective, AVS was not entirely effective. It performed very well in specific tasks, such as detecting high crowd density, wrong-way traffic, or entry into prohibited zones. However, its results were uncertain regarding the detection of weapons or abandoned objects. The results were reportedly unsatisfactory regarding fire outbreaks or detecting people on the ground.
It is also worth noting that we are speaking here of false positives and true positives — situations where the system flagged, rightly or wrongly, a scenario it was programmed to detect. Due to the lack of a uniform methodology, it was impossible to draw conclusions that also factored in undetected situations. Finally, the deployment of these systems required the mobilization of significant state human resources, particularly within law enforcement. A ministry agent's presence was mandatory during the calibration phase.
The parliamentary report's conclusion is therefore mitigated, indicating the need to demonstrate the relevance of these systems before authorizing them under ordinary law.
Algorithmic Surveillance Continues to Expand
The AVS experimentation has been extended until 2027, and subsequently 2030, despite objections, notably from the Constitutional Council and certain opposition MPs. Its scope of deployment has been broadened to include public buildings and public thoroughfares for counter-terrorism purposes. This extension follows a Senate report recommending the prolongation of the tool's trial phase before considering making it permanent. There was therefore no direct continuity between the JOP and the new testing phases currently underway.
The deployment of AVS has also been authorized in private spaces, specifically for detecting shoplifting in retail stores, though still without allowing the use of biometric data analysis.
The Moral Dilemma
By nature, deploying Algorithmic Video Surveillance solutions conflicts with some of our freedoms because it allows the State to expand its scope of intrusion in the name of security. Yet, we can also argue that if these solutions are effective, they enhance our well-being.
They do so at a marginal cost compared to human agents. Deploying 800 cameras allegedly cost only about 900,000 euros during the Olympic period, which remains negligible compared to the total 1.7-billion-euro budget allocated to the event's security. They can also do so indirectly by freeing up personnel, allowing resources to be redeployed on the ground or to other services.
In any case, the rollout of these solutions seems to be progressing continuously across increasingly larger perimeters within a framework that raises questions. Is this renewed experimentation a way to shift the Overton window without opening a genuine debate? Or, conversely, has the fear of technology we held in the past proven to be disproportionate? Are we being hypocritical about our attachment to biometric data given that we authorize FaceID and that facial recognition is already deployed by our border police? If we criticize the fact that experimentation phases are too short, does that automatically imply that we must broaden the scope of AVS usage?
Louis Fourneau, student at 42 Paris, graduate of Sciences Po Paris
Article written in French without AI assistance, translated in English with the help of Gemini AI.
Sources:
Summary report on the use of algorithmic video surveillance during the Olympic Games (French Senate)
CNIL summary on the use of AVS during the Olympic Games (French Data Protection Authority)
Independent media overview of the experimentation (based on the Senate report)