India deploys facial recognition surveilling millions of commuters

Indian politics & policy updates

Sign up to myFT Daily Digest to be the first to know about Indian politics & policy news.

Indian Railways, one of the world’s busiest urban rail systems, has deployed a network of almost 500 facial recognition cameras to track millions of daily commuters, as the Indian government increases its surveillance efforts.

The system, developed by Russian start-up NtechLab, has been live for the past month at 30 railway stations in the densely populated western states of Gujarat and Maharashtra, including the city of Mumbai. The latter’s suburban trains carry more than 7m passengers daily.

Over the past five years, India has been escalating its use of video surveillance, including facial recognition, across the country, with at least 40 government-funded projects tracked by the AI Observatory, an independent research organisation.

The government has also issued an open tender for an integrated nationwide central system called the National Automated Facial Recognition System.

“What’s clear from all these projects is that this government is putting surveillance first,” said Divij Joshi, an independent lawyer and researcher based in Bangalore, who created the AI Observatory. “Often security becomes the go-to excuse for these things, but it is so weak and widely defined with such few safeguards that it could easily turn into moral policing of interfaith couples or activists, or whomever they want.”

Andrey Telenkov, NtechLab’s chief executive, said its system could simultaneously recognise up to 50 people in a single frame, including those wearing medical masks, and would be used to count passenger traffic at any given time on the network. He added that the technology could also identify criminals, track people of interest through live footage, and search for missing persons.

NTech has been linked to several state-backed facial recognition projects in Russia, and has received investment from state defence industry conglomerate Rostec. In Moscow alone, the company has deployed more than 150,000 cameras on behalf of the city government.

“India is a huge market for video surveillance, among the biggest worldwide. This is the entry point to one of the largest clients there,” Telenkov said. “Now it is a few railway stations, but I believe they do want to expand the solution across the entire network.”

Civil society activists have raised alarms about the widespread curtailment of civil liberties in India through the use of biometric systems, including tracking down protesters. Recently, the police used a facial recognition system known as Trinetra to find and arrest more than 1,000 farmers who protested at an anti-government rally. Police have adopted facial recognition technology as a law-enforcement tool despite the absence of a national law to define limits on its use.

Telenkov said this was one of the most technically challenging projects his team had faced, because of the volume of people pouring through the camera’s field of view. However, he said that NTech’s technology had come out on top in a competitive test against half a dozen other systems.

He added that the company had active projects in the Middle East, Africa, Latin America and Russia. “We have shown a very high accuracy on all skin types,” he said.

However, Joshi said that “on the evidence, the capabilities of the systems that are being bought by the government aren’t great at all”.

“From court documents that have been presented, we learnt that the accuracy of these systems in the wild for the Indian demographic is really, really poor. For the missing-child system in place by Delhi law enforcement, the accuracy was less than 1 per cent,” he added.