Pop superstar Taylor Swift had her concert in Paris guarded by AI-powered video surveillance as the city prepares for heightened security risks during the Summer Olympics 2024.

The French police deployed the technology at two Paris metro stations used by fans attending the concert. The decision to deploy the video system was justified by the heightened risk of acts of terrorism, according to Euronews.

The AI system was installed in preparation for the Olympics scheduled in July and August. The algorithms deployed for real-time surveillance detect suspicious or potentially dangerous events, including weapons, fire, bodies on the ground, abandoned packages and abnormal crowd behaviors.

The surveillance cameras are also being tested during other events, including the Cannes Film Festival. The AI surveillance network will watch over movie stars alongside some 40,000 attendees until May 25. Local authorities say they are using 17 AI cameras while the city on the French Riviera already has the “densest video protection network in France” with 884 video cameras, Barrons reports.

AI cameras were also tested during recent concerts by the English band Depeche Mode.

Will Paris’ surveillance system be efficient enough?

While the deployment of the system will be strictly limited to the Olympics, introducing algorithmic tracking has drawn criticism from activists and lawmakers. Some are concerned that the Paris Olympics will set a precedent for other events. But the system also raises other questions, including those related to its efficiency.

Losing footage due to system overload could have significant security implications, according to Vishwa Vijoyendra Narayan, director of product management at Quantum.

“Losing vital sections of footage wouldn’t go down well with the public, heightening concerns about reliability and transparency,” Narayan writes for Infosecurity Magazine.

Legacy network video recorders (NVRs) were not built to support video surveillance operations at scale. Modern surveillance platforms are replacing traditional NVR servers with a single compute and storage software-based platform for video recording and advanced analytics. These allow customers to host more than 10,000 cameras on a single system, he notes.