Paris entered the Olympics under an unusually heavy security presence, with barriers, police patrols, soldiers, QR-code controls near the river Seine, and a new layer of automated monitoring inside parts of the transport network. The most contested change is not the most visible one: algorithms are being connected to CCTV feeds to help identify possible threats in real time.
AI surveillance moves into transport stations
The system is being used in 46 train and metro stations during the Olympics, according to an announcement by the Prefecture du Paris, a unit of the interior ministry. One of those locations is Porte de Pantin, near La Villette, where the Olympics’ Park of Nations lets fans eat or drink in pavilions dedicated to 15 different countries.
At the Porte de Pantin metro station, cameras sit in a gray metal box near the ceiling. A sign nearby tells passengers that they are part of a video surveillance analysis experiment and that RATP, the company that runs the Paris metro, is likely to use automated real-time analysis of CCTV images in which they may appear. The notice says the experiment runs until March 2025.
The technology was first tested in Paris back in March at two Depeche Mode concerts. For the Olympics, it has become part of a broader security operation in a city already divided by 40,000 barriers and guarded by police officers and soldiers.
What the algorithms are allowed to detect
Under a March 2023 law, algorithms can search CCTV footage in real time for eight events. The source lists examples including crowd surges, abnormally large groups of people, abandoned objects, weapons, and a person falling to the ground.
The basic argument for the system is practical. With thousands of cameras, human operators cannot watch every feed closely at all times. Algorithmic CCTV analysis is meant to flag patterns that deserve attention, so a person can review the alert and decide what should happen next.
Matthias Houllier, cofounder of Wintics, describes the technology as a way to transform existing CCTV cameras into a stronger monitoring tool. Wintics is one of four French companies that won contracts to deploy algorithms at the Olympics.
According to Houllier, the system does not make automatic decisions. He says the goal is to draw an operator’s attention, after which people can double-check the situation and choose a response. His team trained interior ministry officials on how to use the company’s software, while the officials decide how it is deployed.
From counting cyclists to monitoring crowds
Wintics’ role in Paris did not begin with Olympic security. The company won its first public contract in Paris in 2020, using algorithms to count cyclists in different parts of the city. That work helped Paris transport officials as they planned to build more bike lanes.
For that project, Wintics connected its algorithms to 200 existing traffic cameras. The system, which is still in operation, identifies and counts cyclists on busy streets. Houllier frames the Olympics work as a natural evolution of that earlier deployment because, in his view, the same underlying technology analyzes anonymous shapes in public spaces.
For the Olympics deployment, Wintics trained its algorithms on open source and synthetic data. The systems can be adapted to count people in a crowd or people falling to the floor, then alert operators once a threshold is exceeded.
Houllier also argues that this approach is different from facial recognition systems used at past global sporting events, including the 2022 Qatar World Cup. He says the company is not analyzing faces, license plates, personal data, or behavioral analytics, but looking at shapes.
Rights groups see a broader risk
Privacy activists reject the claim that this form of AI surveillance protects personal freedoms. Noémie Levain, a member of La Quadrature du Net, argues that analyzing images of people still means processing personal and biometric data. Her group opposes AI surveillance and is distributing 6,000 posters in Paris warning about algorithmic surveillance and the capture of public spaces.
Levain’s concern is not limited to what happens during the Games. She worries that once the infrastructure exists, it will remain after the athletes leave. In her view, the system expands the reach of police and security services across the city and could reinforce existing patterns of discrimination.
That fear sits beside a familiar Olympics debate. The source notes that complaints about extreme security have followed previous Games, with labels such as Lockdown London, Fortress Tokyo, and the arms race in Rio. In Paris, however, the dispute has shifted toward software that many passengers may never notice unless they stop to read a sign.
The future question behind the Paris experiment
The Paris deployment points to a larger issue for cities: what happens when CCTV networks become too large for people to monitor directly. Supporters see AI surveillance as a way to make existing cameras more useful during crowded, high-risk events. Critics see it as a step toward routine automated oversight of public life.
That tension is why the Olympics matter beyond the sporting event itself. The system is presented as a temporary experiment tied to transport stations, security pressure, and real-time alerts. But opponents argue that temporary security tools can become permanent parts of city infrastructure.
For now, the experiment runs through March 2025, and city representatives did not reply to questions on whether there are plans to use AI surveillance outside the transport network. The unanswered question is what Paris will keep when the Games are over: only the memory of an intense security operation, or a new model for watching public space.