Paris entered the Olympics under an unusually dense security presence: barriers across the city, police patrols, soldiers in public areas, and access restrictions around the river Seine. Alongside the visible controls, a quieter experiment is drawing some of the sharpest scrutiny: AI surveillance connected to CCTV cameras in the transport network.
The system does not sit at the center of public attention in the way roadblocks or patrols do. That is part of the controversy. It works in the background, scanning camera feeds for situations that authorities have defined as possible security concerns.
What Paris is deploying
The AI system is being used to analyze CCTV footage in real time at train and metro stations. According to an announcement by the Prefecture du Paris, a unit of the interior ministry, the technology is being deployed at 46 train and metro stations during the Olympics.
One site described in the source is Porte de Pantin metro station, near La Villette, where the Olympics’ Park of Nations is located. A sign at the station tells passengers they are part of a “video surveillance analysis experiment” and says the company that runs the Paris metro, RATP, “is likely” to use “automated analysis in real time” of CCTV images “in which you can appear.” The experiment runs until March 2025.
Under a March 2023 law, algorithms may scan CCTV footage in real time for eight “events.” The examples given include crowd surges, unusually large groups of people, abandoned objects, weapons, or a person falling to the ground.
The system was first tested in Paris back in March at two Depeche Mode concerts. During the Olympics, it becomes part of a much larger security environment, where authorities are already managing crowds, closures, vetting, and transport pressures.
Why officials and vendors say it is needed
The argument for AI-assisted CCTV is straightforward: modern events generate more video than people can reasonably watch. Cameras may be everywhere, but human attention is limited. Supporters say software can help operators notice possible incidents faster.
Matthias Houllier, cofounder of Wintics, describes the technology as a way to make existing cameras more useful. Wintics is one of four French companies that won contracts to have algorithms deployed at the Olympics. “What we're doing is transforming CCTV cameras into a powerful monitoring tool,” he says. “With thousands of cameras, it's impossible for police officers [to react to every camera].”
Wintics had already worked with Paris before the Olympics. In 2020, it won its first public contract in the city, using algorithms connected to 200 existing traffic cameras to identify and count cyclists in busy streets. That data was used to help transport officials plan more bike lanes, and the system is still in operation.
For the Olympics, Houllier presents the shift from transport planning to security as a technical continuation. “The technology is the same,” he says. “It's analyzing anonymous shapes in public spaces.”
What the algorithms are supposed to do
Wintics trained its algorithms on open source and synthetic data. For the Olympic security use case, the systems can be adapted to count people in a crowd or count people falling to the floor, then alert operators when a threshold is crossed.
Houllier emphasizes that the system is not meant to act on its own. “That's it. There is no automatic decision,” he says. His team trained interior ministry officials to use the software, and those officials decide how it is deployed. The goal, as he puts it, is to draw an operator’s attention so a person can check the situation and decide what to do.
He also argues that this approach is different from facial recognition. In his view, AI analysis of shapes can support event safety without putting personal freedoms at risk. “We are not analyzing any personal data. We are just looking at shapes, no face, no license plate recognition, no behavioral analytics.”
Why critics see a deeper risk
Human rights groups have denounced the security measures as creating “unacceptable risks to fundamental rights.” For privacy activists, the issue is not only what the software is designed to detect, but what it means to normalize algorithmic monitoring in public spaces.
Noémie Levain, a member of La Quadrature du Net, which opposes AI surveillance, describes the software as an expansion of police visibility. “The software is an extension of the police,” she says. “It's the eyes of the police multiplied.”
Levain rejects the claim that the technology avoids personal data simply because it does not identify faces or license plates. “When you have images of people, you have to analyze all the data on the image, which is personal data, which is biometric data,” she says. “It's exactly the same technology as facial recognition. It's exactly the same principle.”
Her concern also extends beyond the Olympics. La Quadrature du Net has prepared 6,000 posters warning Parisians about “algorithmic surveillance” and urging them to resist the “authoritarian capture of public spaces.” Levain worries that systems introduced for a major event will remain after the athletes leave.
The bigger question after the Games
The Olympics have often been accompanied by intense security measures, and the source notes that past events have been described with terms such as Lockdown London, Fortress Tokyo, and an “arms race” in Rio. Paris fits that pattern, but the AI layer changes the debate because it can be harder for the public to see and assess.
French interior minister Gérald Darmanin has defended the broader security posture as vigilance. He told reporters that France faces the “biggest security challenge any country has ever had to organize in a time of peace.” He has also said “potentially dangerous individuals” were found applying to work or volunteer at the Olympics, including 257 radical Islamists, 181 members of the far left, and 95 from the far right. He told BFM that a Russian citizen had been arrested on suspicion of plotting “large scale” acts of “destabilization” during the Games.
Those facts help explain why officials want tools that can help manage risk. They do not settle the civil liberties dispute. The core question is whether AI surveillance remains a narrow response to a temporary event or becomes a lasting part of city life.
Levain’s warning is that the Olympics provide a reason to accelerate a system that government, companies, and police may continue using afterward. “The Olympics is an excuse,” she says. “They—the government, companies, the police—are already thinking about after.”