Indian Temples Are Scanning Visitors’ Faces

The Hindu deities in some of India’s largest temples now recognize millions of worshippers by their faces—not because of divine powers, but thanks to technological ingenuity.

The controversial Ram Mandir in Ayodhya, Mahakaleshwar Temple in Ujjain, and Venkateswara Swami Temple in Tirupati—located in north, central, and south India, respectively—have installed hundreds of cameras equipped with facial recognition technology, capturing the personal information of the tens of thousands of devotees who visit these temples every day. While the initiative has been touted as a technological marvel, critics fear that it could be a testing ground for mass surveillance and data harvesting in the absence of relevant laws in India.

At the Venkateswara Swami Temple, devotees are photographed at the entrance, and their information is matched with the details provided during registration for the visit and securing accommodation. This is managed by the Tirumala Tirupati Devasthanam (TTD), an independent trust that oversees the temple. Rolled out last year, the facial recognition system intends to restrict entry for those with a criminal history and eradicate the rampant use of middlemen extorting devotees, according to L.M. Sandeep, TTD’s general manager for information technology. These middlemen often charge visitors anywhere between a few hundred and several thousand rupees (about $5 to $50) for quicker access to worship the temple’s main idol and promises of guaranteed affordable trust-run accommodation. And they’re pervasive: Local police arrested 16 of them for selling such services in 2021.

Sandeep told Foreign Policy that the temple “only captures photos” and that any data provided by the devotees is “deleted after a certain time.”

However, experts aren’t convinced. They argue that facial recognition technology is inaccurate and flawed in design and operation. It is also unclear where the data collected at temples is stored and used.

“These seem like surface-level excuses,” Disha Verma, an associate policy counsel at digital rights organization Internet Freedom Foundation, told Foreign Policy. “Facial recognition in its current state is not accurate. Maintaining a religion-based database can have catastrophic effects for minority groups.”

India has quickly adopted facial recognition technology since 2018, when Delhi police first started using the system to find missing children. The Internet Freedom Foundation has tracked 170 facial recognition technology systems installed across India, with the highest placements in Maharashtra state, followed by Delhi, Telangana, and Gujarat.

The technology is used in both public and private spaces, including cafes and schools, though its most insidious use has come from law enforcement officials.

A 2021 report by the Vidhi Center for Legal Policy, a Delhi-based independent think tank, noted that the use of facial recognition technology (FRT) by police raises questions involving efficiency, surveillance, and discrimination, adding that the organization mapped usage in different Delhi police jurisdictions and found that “Muslims are more likely to be targeted by the police if FRT is used.”

The technology has also been deployed to arrest dissidents in India. In 2021, the Delhi police commissioner said that facial recognition technology helped identify more than 100 people protesting against the Citizenship Amendment Act the previous year. The Islamophobic act—which offered a path to citizenship for undocumented immigrants from Afghanistan, Bangladesh, and Pakistan, but only for non-Muslims—led to protests and riots in northeast Delhi in February 2020, killing at least 53 people.

Ameen Jauhar, a lawyer specializing in data governance and artificial intelligence regulation, told Foreign Policy that facial recognition technology has been used to target and identify vulnerable populations. However, there are significant error margins—even technologies claiming to be highly accurate often provide only 70 percent to 80 percent accuracy.

“Relying on a tech that has the potential to result in deprivation of right and liberty, and doing that without any kind of legislation or legal checks and balances, is unconstitutional,” he said. “The gaps in law and legislation are facilitating arbitrary and unbridled use of facial recognition technology.”

Earlier this year, Indian Prime Minister Narendra Modi consecrated the Ram Mandir in Ayodhya, a giant, $225 million temple honoring the Hindu deity Rama. The building replaced a 16th-century mosque that was razed by a violent Hindu mob in 1992. In 2019, India’s Supreme Court allowed the building of a new temple in a ruling after almost three decades of land dispute between the region’s Hindus and Muslims—creating a site that has now become a locus for Hindu nationalists

Modi called the temple’s opening the start of a “new era,” while his Hindu nationalist Bharatiya Janata Party-led (BJP) government wrote that Ayodhya was “set to become the world’s first artificial intelligence (AI)-powered Vedic city.”

That vision was tested in Ram Mandir during its consecration on Jan. 22. Indian start-up Staqu Technologies was roped in to strengthen security around the temple, powered by its AI-equipped JARVIS video analytics software.

In a blog post, the company said it collaborated with police in Uttar Pradesh, the state where Ayodhya is located, to identify “suspicious faces” based on the 800,000 criminals registered in a law enforcement database. Staqu also mentioned that it uses reverse facial recognition and attribute-based searches, allowing users to identify people based on their clothing, color, and accessories. ​

“We piloted a dystopian advanced technology not just because it’s a temple but [because it’s] a mass gathering of people,” said Verma, of the Internet Freedom Foundation. “The attribution-based facial recognition and close monitoring happened because it’s such a huge gathering and an experiment in a high degree of surveillance.”

Verma noted that such practices are taking place in a highly charged sociopolitical climate that uses religion to leverage votes during elections—and that the data could be used for surveillance and targeted voter campaigning.

Many political parties in India make use of religion, and BJP leaders especially have resorted to Islamophobic rhetoric during election campaigns to appease Hindu voters.

“Keeping a database of Hindus and frequent temple visitors is a highly sought-after dataset,” Verma said. “Tomorrow, political parties can target Hindus and Hindu votes, and the reverse can happen in minority places of worship.”

The Mahakaleshwar Temple in Ujjain, which sees up to 200,000 devotees daily, installed AI-enabled facial recognition technology for crowd management in July 2023. The temple’s administrator told the Print that the system generates a profile for every person entering the temple, helping the temple’s managers to track the number of people inside.

Last year, the government of Karnataka, a coastal southwestern state, also introduced a scheme in which devotees from the state visiting the Kashi Vishwanath Temple in the holy city of Varanasi—where Modi recently won the constituency for the third time—could receive a 5,000 rupee ($60) cash subsidy by uploading a geotagged selfie in a government portal.

“You don’t know what they are doing with these photos,” Verma said. “You are relinquishing the right to privacy and your face just to get the benefit the state owes you. For elections or government schemes, religion-based surveillance can lead to the surveillance of minorities. If you can put it in temples, then you can use it in mosques, too.”

In August 2023, India’s parliament enacted the country’s first data protection law—but it’s yet to be enforced, thus leaving individual privacy in a de facto limbo, said Jauhar, the lawyer. He argued that the law does not protect people from surveillance, and noted that there are no rules or institutional infrastructure under the Digital Personal Data Protection Act to govern emerging technologies such as the facial recognition system.

Even countries that make heavy use of facial recognition systems—such as China—are gradually drafting laws to limit private usage and protect data. Last year, the Cyberspace Administration of China released draft rules detailing the collection, storage, and usage of facial recognition data, while prohibiting certain facilities—including hotels, banks, and airports—from forcing individuals to use the technology. The rules also said that consent must be obtained from individuals, and the collected images and personal information must not be used for other purposes without consent.

Though India’s data protection law also has provisions on consent, saying it should be “free, specific, informed, unconditional, and unambiguous,” analysts say those implementing the facial recognition technology are unlikely to follow the guidelines unless they are held accountable.

Gagandeep Kaur, an associate professor of law at the University of Petroleum and Energy Studies in Dehradun, told Foreign Policy that many places using facial recognition in India do not ask for explicit consent from individuals, though they often have signs noting that the area is under surveillance. She added that consent is usually implied, especially in places such as temples.

“This is a gross invasion of privacy and may lead to misuse of data,” said Kaur, who has written about the legal implications of facial recognition technology in India. “It is a modern approach to managing security and streamlining operating in religious settings where people are unlikely to challenge or question its use.”

However, few individuals have challenged facial recognition technology in courts over the past years, although the Internet Freedom Foundation has provided legal support in some cases.

A prominent social activist filed a landmark petition against the technology in the southern state of Telangana after police officers asked him to remove his mask during a COVID-19 surge and took his photo while returning from work in 2021, which he said was unlawful and done without any measures to prevent misuse.

In another case, a resident from Chennai challenged the deployment of facial recognition technology in the southern state of Tamil Nadu, where Chennai is located, arguing that the technology has been installed through “covert and non-consensual means” and that there is little or no information on how such data is being stored and used.

The high courts in Telangana and Chennai, where the petitions were filed, are yet to rule on the cases.

The lack of laws and limited information on the technology have made it challenging for the public to grasp the full extent of its deployment.

“India is callously treating facial data, and the law reflects our attitude toward it,” Verma said. “We are at a stage where surveillance is normalized and pervasive in all factors of life. We urgently need surveillance reforms and red lines, as well as laws that address this and do not skimp over the obligation of the state and the government.”