Inside Safe City, Moscow’s AI Surveillance Dystopia

Moscow promised residents lower crime rates through an expansive smart city project. Then Vladimir Putin invaded Ukraine.
A surveillance camera in Red Square Moscow with the Kremlin in the background
Photograph: Andrey Rudakov/Bloomberg/Getty Images

Sergey Vyborov was on his way to the Moscow Metro’s Aeroport station last September when police officers stopped him. The 49-year-old knew that taking the metro could spell trouble. During a protest against Russia’s invasion of Ukraine, police had fingerprinted and photographed him. He’d already been detained four times in 2022. But he was rushing to his daughter’s birthday, so he took a chance.

Vyborov wasn’t arrested that day, but the police informed him that he was under surveillance through Sfera, one of Moscow’s face recognition systems, for participating in unsanctioned rallies. Considered one of the most efficient surveillance systems, Sfera led to the detention of 141 people last year. “Facial recognition, and video cameras in general in a totalitarian state, are an absolute evil,” Vyborov says. 

Vyborov finds himself at the bottom of a slippery slope that privacy advocates have long warned about. Under the guise of smart city technology, authoritarian and democratic governments have rolled out huge networks of security cameras and used artificial intelligence to try to ensure there is no place to hide. Cities have touted the ability of such systems to tackle crime, manage crowds, and better respond to emergencies. Privacy campaigners say such systems could be used as tools of oppression. In Moscow, Vyborov and countless others now face that oppression on a daily basis.

The Russian capital is now the seventh-most-surveilled city in the world. Across Russia, there are an estimated 21 million surveillance cameras, and the country ranks among the top in the world in terms of the number of connected surveillance cameras. The system created by Moscow’s government, dubbed Safe City, was touted by city officials as a way to streamline its public safety systems. In recent years, however, its 217,000 surveillance cameras, designed to catch criminals and terrorists, have been turned against protestors, political rivals, and journalists. 

“Facial recognition was supposed to be the ‘cherry on top,’ the reason why all of this was built,” says a former employee of NTechLab, one of the principal companies building Safe City’s face recognition system.

Following Russia’s invasion of Ukraine, Safe City’s data collection practices have become increasingly opaque. The project is now seen as a tool of rising digital repression as Russia wages war against Ukraine and dissenting voices within its own borders. It is an example of the danger smart city technologies pose. And for the engineers and programmers who built such systems, its transformation into a tool of oppression has led to a moment of reckoning. 

An Innocent Start

Founded in 2015, NTechLab caught the attention of the global press with the February 2016 launch of FindFace, an app that allowed anyone to identify faces by matching them with images gathered from social network VKontakte, Russia’s Facebook equivalent. Met with warnings of the “end to public anonymity,” the app was reportedly downloaded by 500,000 people within two months of its launch. But for NTechLab, it was primarily a proof of concept for its nascent face recognition algorithm.

NTechLab still felt like a startup when one former employee, who asked not to be named for privacy reasons, joined the company. And he was drawn in by the complexity of the work.

“From [an] engineering point of view, it’s very interesting to work with: It’s very difficult,” he says. 

After the release of FindFace, NTechLab began selling its face recognition tech to small businesses, such as shopping malls that could use it to catch shoplifters or see how many people return to certain stores. But NTechLab was also working with the Moscow Department of IT Technology (DIT), the government department tasked with building Moscow’s digital infrastructure. In 2018, when Russia hosted the FIFA World Cup, NTechLab’s face recognition tech was connected to more than 450 security cameras around Moscow, and its tech reportedly helped police detain 180 people whom the state deemed “wanted criminals.”

At its inception, Moscow’s face recognition system was fed official watchlists, like the database of wanted people. The system uses these lists to notify the police once a person on the list is detected, but law enforcement can also upload an image and search for where a person has appeared. Over the years, security and law enforcement agencies have compiled a database of the leaders of the political opposition and prominent activists, according to Sarkis Darbinyan, cofounder of digital rights group Roskomsvoboda, which has been campaigning for a suspension of the technology. It remains unclear who is in charge of adding activists and protesters to watchlists.

In March 2019, following the success of the World Cup trial—some of Russia’s “most wanted” people were arrested while trying to attend matches—the Moscow Department of Transportation, which operates the city’s metro, launched its own surveillance system, Sfera. By October 2019, 3,000 of the city’s 160,000 cameras were enabled with face recognition tech, according to interior minister Vladimir Kolokoltsev.

NTechLab was one of many companies building the slew of systems that would later be branded Safe City. International companies, from US tech firms such as Nvidia, Intel, and Broadcom to South Korea’s Samsung and Chinese camera maker Hikvision, worked alongside local firms such as HeadPointNetris, and Rostelecom that have developed various components of the surveillance systems. According to procurement documents cited by the UK’s BBC, three companies besides NTechLab created face recognition tech for Moscow’s growing surveillance apparatus, including Tevian, and Kipod, and VisionLabs. Moscow's Transportation Department said in social media posts that Sfera was built using VisionLabs technology, although the company downplays its involvement.

NtechLab says it operates in compliance with local laws and does not have access to customer data or camera video streams. Nvidia and Intel say they left Russia in 2022, with Nvidia adding that it does not create software or algorithms for surveillance. Broadcom and Samsung also say they stopped doing business in Russia following the invasion. VisionLabs says it only provides the Moscow Metro with its face recognition payment system. Other companies did not respond to requests for comment. The DIT and the Moscow Department of Transportation did not respond to requests for comment.

At the end of 2018, as Russia cracked down harder on political dissent online and in the streets, the DIT started to change, says a former employee who asked to remain anonymous for safety reasons. The department used to just be the “technical guys” providing assistance to security services, with the Moscow government recruiting highly paid IT specialists to make the most efficient systems possible, according to Andrey Soldatov, an investigative journalist and Russian security services expert. But according to the former employee, the DIT was beginning to reflect the Kremlin’s authoritarian bent.

Then came Covid. 

Safe City launched in 2020, at the height of the Covid-19 pandemic. Russia, like some other countries, seemingly used the pandemic as grounds to expand its surveillance systems to catch people breaking self-isolation rules. By mid-March 2020, Safe City’s face recognition system had caught 200 people breaking lockdown restrictions. At the same time, Moscow introduced a regulatory sandbox for the development of AI applications with the participation of large IT companies, exempting authorities from the country’s already lax data protection requirements. “With Covid, [the DIT] essentially became a part of the repressive apparatus,” says Soldatov.

In addition to its network of more than 200,000 cameras, Safe City also incorporates data from 169 information systems, managing data on citizens, public services, transportation, and nearly everything else that makes up Moscow’s infrastructure. This includes anonymized cell phone geolocation data collection, vehicle license plate recognition, data from ride-hailing services, and voice recognition devices. As Safe City was still rolling out in 2020, the Russian government announced plans to spend $1.3 billion deploying similar Safe City systems across Russia. From the outside, the potential for the system to be abused seemed obvious. But for those involved in its development, it looked like many other smart city projects. “No one expected that the country would turn into hell in two years,” says one former NTechLab employee, who asked to remain anonymous for safety reasons.

The Black Box Problem

Attempts to break open Moscow’s digital black box have been stonewalled. Alena Popova, whose image was captured during a protest against politician Leonid Eduardovich Slutsky in April 2018, filed the first lawsuit against Moscow’s DIT for allegedly violating her privacy, seeking a ban on face recognition tech. The case was thrown out, but Popova has continued to file lawsuits, including one at the European Court of Human Rights—which Russia is no longer a part of. 

While Moscow operates one of the world’s most pervasive surveillance systems, Russian law does not safeguard individual privacy. With seemingly no hope of recourse, some activists have been forced to leave Russia altogether. Popova is now on the list of foreign agents and is living in an undisclosed overseas location. “I will not apply to any political asylum in any country because I would like to go back to my own country and fight back,” she says.

A key concern is that Moscow’s surveillance system was designed to conceal its data collection from Moscow’s 12 million residents, says Sergey Ross, founder of the Collective Action Center think tank and a former Moscow politician. Although the system is run by the Moscow government, elected members of the Moscow City Duma say they are excluded from regulating face recognition systems and have little insight into how it is being used. “It’s a complete black box,” says Ross.

“It was clear that sooner or later the technology would be used to catch activists and dissenters,” says Roskomsvoboda’s Darbinyan. 

Russia made almost 20,500 political arrests in 2022, according to data from human rights media organization OVD-Info, which characterizes the number as “unprecedented.” The arrests have sparked fears that Safe City will be expanded to catch draft dodgers—although former NTechLab employees say that doing so would be technically difficult to pull it off because of too many false positives. Still, Moscow police appear to be using face recognition to aid Russia’s war efforts in other ways.

In September 2022, just after Putin announced additional mobilization for the war against Ukraine, Viktor Kapitonov, a 27-year-old activist who’d protested regularly since 2013, was stopped by two police officers after being flagged by face recognition surveillance while he approached the turnstiles in Moscow’s marble-covered Avtozavdodskaya metro station. The officers took him to the military recruitment office, where around 15 people were waiting to enlist in Putin’s newly announced draft. 

“They let me in without waiting in line as if I were some sort of VIP person,” he says. The recruiters wanted to force Kapitonov to enlist, but he ended up escaping the draft. “I explained that I am not fit, I have a disability.”

A Guilty Conscience

From 2017 to 2020, NTechLab became one of Russia’s fastest-growing companies. Other face recognition firms have cashed in as well: The revenue of Russian face recognition developers grew between 30 and 35 percent in 2022, thanks in part to deals struck in the Middle East, Southeast Asia, India, and South America. Russia’s national AI strategy has supported such firms with grants, tax exemptions, and subsidies, which have benefited both startups and state corporations, including tech and finance giant Sber, telecom provider Rostelecom, and defense firm Rostec, which previously owned a minority stake in NTechLab. While NTechLab continues to work globally, reporting a revenue increase of 35 percent in 2022, it has also faced a backlash against its work with the Russian state.

In June of last year, a “name-and-shame” list of NTechLab employees was published [in Russian] with information collected from social media. The project went viral, and some employees reported being harassed online. Artem Zinnatullin, a software engineer now based in the US, says he published the list after NTechLab sold its new silhouette recognition technology to the Moscow government in June 2022. To him, it signaled support for Russia’s war in Ukraine. In the post, he called NTechLab “the blacksmith of the Digital Gulag.” Zinnatullin, who says he knew people arrested with the help of face recognition technology, believes publishing the list of NTechLab employees was only fair. “You recognize people on the street, it’s only fair if we use public data to recognize who you are,” he says.

Unlike many face recognition companies that keep a low profile, NTechLab’s splash with FindFace has turned it into a recognized brand. Employees say this high profile has made them into scapegoats. 

As arrests of activists and politicians mounted, the ethics of NTechLab’s technology became a recurring topic at company meetings. NTechLab staff have resisted the use of the company’s face recognition in rallies and refused to sell the technology to the military, according to people familiar with these discussions. Still, the NTechLab leadership concluded that the technology was ultimately positive—even if the occasional dissenting voice was arrested because of it. 

“We all saw these positive examples, we saw how it really catches criminals,” says one former NTechLab employee. “Most people in NTechLab would say they were doing something very good, technologies that can help and save people’s lives. It really did.”

As Russia furthered its march toward authoritarianism in 2021, NTechLab leadership began talking about moving the company abroad, according to people familiar with internal company discussions. But with lucrative government contracts abounding—NTechLab received a $13 million investment from the Russian Direct Investment Fund, the country’s sovereign wealth fund, in September 2020—its investors resisted the idea. The company was also changing. Its founders, Alexander Kabakov and Artem Kukharenko, stepped down from NTechLab—and both left Russia in December 2021 and February 2022, respectively, declaring their anti-war stance on social media. 

Other employees left amid an exodus of IT talent from Russia. The war changed how they viewed their work. “Looking back, we realize that we shouldn’t have done it,” says an NTechLab employee. “But even in 2017 and 2018, it was a completely different country. At least, that’s how it seemed to those who weren’t very immersed in politics.”

A Growing Threat

Russia’s Safe City projects show no sign of slowing. As more surveillance systems are deployed across the country, Moscow’s DIT is planning to centralize video streams collected across all regions into its own system. And new projects to digitize public services may make it even easier for the government to eventually create large databases where everyone can be found, according to Popova. “It is really scary,” she says. “If they digitalize all the databases and combine them to make this joint database, they can find everybody.” In July, Putin signed a federal law that funnels personal biometric data collected in the country into a single system—an effort to obtain an “almost unlimited monopoly” on the collection and storage of biometrics, says Roskomsvoboda’s Darbinyan. 

In a further expansion of the Safe City project, Rostec is also reportedly developing software that will help authorities predict riots and prevent their escalation by analyzing media reports, data from social networks, video cameras, and other sources. Rostec did not respond to a request for comment on its development of these systems.

Similar systems have been developed in some Chinese cities, and Russia is now playing catch-up. “The Russian government would probably like to move toward China, but they do not yet have the necessary technology,” says Kiril Koroteev, head of international practice at the Russia-based Agora International Human Rights Group.

For now, many activists in Russia are left to do whatever they can to skirt the country’s growing surveillance apparatus, including avoiding the Moscow Metro. Kapitonov hopes that a balaclava will keep him safe, while Vyborov aims to ride the metro early in the morning, when there are fewer police around to detain him. 

“I think that it was inevitable that such a system would be made sooner or later,” says one former NTechLab employee. Face recognition is like a knife, he says: It can be used to cut food, but it can also be used to cut innocent people. He now regrets that NTechLab played a key role in building Moscow’s Safe City project. He has left Russia and doesn’t think he will work on face recognition again. “I do not want to mess with it anymore,” he says.

Update 9:25 am ET, February 6, 2023: Clarified the role of VisionLabs in the Sfera system and that NTechLab's founders have since left the company.