Paris Olympics crowd scanning fuels AI surveillance fears

London –

France's plans to use artificial intelligence to scan thousands of athletes, coaches and spectators gathering in Paris for the Olympics are a form of creeping surveillance, rights groups have said.

In recent months, French authorities have been testing artificial intelligence surveillance systems in locations including train stations, concerts and football matches.

When the Olympics open in late July, these systems will scan spectators, check abandoned luggage, detect weapons, and do a variety of other things.

French authorities have said the tools will not be fully operational before the Olympics, but will be used afterwards by police, fire and rescue services, as well as some French transport security agencies, until March 31, 2025.

Activists fear that AI-enabled surveillance could become the new normal.

“The Olympics are a perfect opportunity to test this kind of surveillance under the guise of a security issue, paving the way for even more intrusive systems like facial recognition,” said Katia Roux, advocacy leader at Amnesty International France.

Station and Taylor Swift

The French government has enlisted four companies in the initiative: Videtics, Orange Business, ChapsVision and Wintics.

The companies' security platforms measure eight key indicators: traffic moving against the flow, people in restricted areas, crowd movement, abandoned luggage, presence or use of weapons, overcrowding, bodies on the ground and fires.

Concerts by Depeche Mode and the Black Eyed Peas, as well as soccer matches between Paris Saint-Germain and Olympique Lyonnais, have served as testing grounds for the software.

Further testing was carried out on crowds passing through Nanterre-Préfecture and La Défense Grande Arche stations to see Taylor Swift, as well as 40,000 attendees at the Cannes Film Festival.

Cannes Mayor David Lissard said his city already has “the most dense video surveillance network in France” with 884 cameras – one for every 84 residents.

According to a 2020 report, there are around 90,000 video surveillance cameras across France monitored by police and gendarmerie.

“While most of these use cases may not seem to involve identifying or profiling individuals, the big concern is that they still require the deployment of surveillance infrastructure and could enable the most invasive mass surveillance with a simple software update,” said Daniel Rufer, senior policy analyst at digital rights group AccessNow.

“The public will have little to no oversight over what these systems are monitoring, what updates are being made, etc., and so will suffer the chilling effect that inevitably results from this kind of public oversight,” he said.

The Olympics will be an AI playground

French lawmakers have tried to assuage criticism with a ban on facial recognition, which authorities say is a red line that must not be crossed.

Wintix co-founder Matthias Houllier said the experiment was “strictly limited” to the eight use cases defined by law, and that features such as crowd movement detection could not be used for other processes such as gait detection, which can identify people from the characteristics of how they walk.

Houllier said that by design, it would be “absolutely impossible” for end users or senior engineers to use Wintics for facial recognition.

Representatives for Videtics, Orange Business and ChapsVision did not respond to requests for comment.

Experts are concerned about how the government will measure the success of these tests and that exactly how the technology works has not been made public.

“There's nowhere near the level of transparency that we need around these technologies,” Rufer said. “There's a very unfortunate claim, particularly in law enforcement and public safety, that we can't have transparency around these systems, which is nonsense.”

“The use of these surveillance technologies, particularly in the areas of law enforcement and public safety, arguably has the greatest potential for harm and therefore demands the highest level of public accountability,” he said.

Privacy activists argue that exceptions to the law would allow “competent authorities” to deploy facial recognition technology for purposes such as national security or immigration.

“This is not a ban. It's actually a permission for law enforcement. People are under the illusion that it's OK because it says this technology is banned except in this situation, this situation, this situation, but those are the situations that are the most problematic,” Lu said.

France's past surveillance activities have also raised concerns: Last November, the nonprofit group Disclosure revealed that law enforcement agencies had been secretly using facial recognition software from Israeli company BriefCam since 2015.

French politicians have suggested there remains a gap between the promise of AI surveillance and its capabilities.

“AI-based video surveillance would not be optimal for the Olympics, but the Olympics would be a great place to test it,” Senator Agnes Kanaier said.

“We will need more internal security forces and private security forces to make up for shortcomings in technology,” she said.

The Interior Ministry, which oversees French law enforcement, did not respond to a request for comment.

In a list of proposals for the future of AI-powered surveillance, the government's Law Committee recommended that the technology should continue on an “experimental basis”, that images captured by the system should also be stored for an extended period and that the equipment should be tested “throughout the entire season” and during smaller events.

“That's why we decided to launch a campaign now to raise awareness about facial recognition technology, even if it won't be used during the Olympics,” Lu said. “If we wait until it's actually used, it will be too late.”