On July 20, thousands of people gathered at Jantar Mantar in Delhi for demonstration. During this time, a journalist saw something in the police surveillance system which he did not expect.
Srishti Jaiswal, an investigative journalist, was reporting at Jantar Mantar. While taking pictures and videos of the protest site, when he zoomed in on a surveillance van parked there, he saw his own face on the monitor inside the van.
According to Srishti, the camera was zooming in and zooming out on her face.
She says, "It took me a second to recognize myself. Then I thought, oh my God, this is me."
He says that he saw different types of monitoring happening on many screens inside the van. He asked the officer present inside the van about this.
The officer only told that there are PTZ cameras installed there, which zoom automatically. When he asked the question again, the door of the van was closed.
There were not only surveillance vans at Jantar Mantar that day. CCTV cameras, drones and police photographers were also present there.
Another journalist Ila Kazmi also recorded pictures and videos of a surveillance van. According to him, 'Live Facial Recognition System' was written on the van.
Their videos show photographs and names of protesters along with recordings. According to him, on some screens, along with the facial image, 'Match Similarities' and its percentage are also visible.
If we understand facial recognition technology in simple words, the camera first captures a picture of a person's face. The software creates a digital copy of that face and then tries to match it with photographs in a database.
But the question of which software was actually used in the system used at Jantar Mantar and with which database the faces were matched, is still not clear.
Journalists Srishti and Kaushik Raj had questioned Delhi Police on this issue for their reporting.
According to him, "The police were asked what technology is being used in the surveillance van, where the data of the recorded people is kept and what is done with it next."
He did not get answers to these questions.
Digital rights researcher Disha Verma also says that one of the biggest problems with the facial recognition system is its opaqueness.
According to him, it is not clear which database the police uses to identify a person.
She raises the question that if the database is local then how will the protesters from other states be identified? And if a national level database is being used then the question of possibility of wrong matching becomes bigger.
We also studied government documents, RTI responses, court records and old reporting to understand the surveillance system seen at Jantar Mantar.
The use of facial recognition in policing in India has increased in the last few years.
In the year 2019, the National Crime Records Bureau (NCRB) had issued a tender for the National Automated Facial Recognition System (AFCS).
Its purpose was to provide such a system of facial recognition to law enforcement agencies, which could help in identification through photographs.
In 2020, the government also gave information about AFCS in the Parliament.
According to the government, it was to be used to help identify criminals, unidentified bodies and missing or found people.
Surveillance infrastructure in cities also increased in the last decade.
In the Integrated Command and Control Centers (ICCC) created under the Smart Cities Mission, a system was developed to monitor different cameras and video feeds of the cities from one place.
Arrangements have also been made to use technologies like video analytics in these centres.
This does not mean that every ICCC uses facial recognition. But these systems have created an infrastructure to simultaneously monitor different surveillance feeds across cities.
We also saw that Delhi Police had previously talked about using facial recognition technology in other cases.
According to Delhi Police statements and information received from RTI, facial recognition technology was used in the investigation of the 2020 North-East Delhi riots and other cases.
In March 2020, the then Home Minister Amit Shah had said in Parliament that during the investigation of Delhi riots, 1,922 people were identified with the help of facial recognition software.
But questions have also been raised on accepting a digital identity as final evidence.
The court had raised questions on the digital identity of the police regarding the identity of Mohammad Arif in a case related to Delhi riots.
In his case, the Delhi High Court held that proving his role on the basis of digital identity is a matter for the subsequent trial process. And at that stage it was not sufficient grounds to keep him in jail.
According to journalist Kaushik Raj, "This case is an example of the fact that there can be a difference between a person's digital identity and his real identity."
The biggest question with facial recognition is not just how accurately the system can recognize a person.
The question is also what will happen if the system goes wrong?
Digital rights researcher Disha Verma gives the example of leaked data from a facial recognition database in Telangana.
According to them, "The leaked system screenshots showed a number of potential matches when searching for a person's photo, showing a high percentage of similarity, even though the faces in the photos did not look alike."
According to him, in such technology the effect of misidentification is not limited to just one percent of what is visible on the screen.
If a person is considered a criminal, he may have to appear before the police, his name may be recorded in some record and further investigation may be affected.
This is why experts ask what are the rights and procedures to challenge that decision when an algorithm marks a person as a suspect or criminal.
The impact of surveillance at Jantar Mantar was not just technical.
According to journalist Srishti Jaiswal, the presence of police photographers was also constantly felt at the protest site.
At the same time, journalist Ila Kazmi says that when she came to know that her face could also be easily identified, she feared that some further action might be taken against her.
Some of the protesters also said they covered their faces because they feared they could be identified through the cameras.
This is where another question of surveillance arises that can the mere fear of participating in a protest stop people from exercising their constitutional rights?
In many parts of the world, strict rules have been made regarding real-time facial recognition in public places.
Digital rights groups have consistently raised questions about the lack of a dedicated comprehensive legal framework for police use of facial recognition in India.
In the RTI responses received by the Internet Freedom Foundation, Delhi Police was asked questions related to the use of facial recognition, data retention and privacy impact assessment.
In these responses, Delhi Police did not provide any clear data retention policy, privacy impact assessment or specific SOP that clearly regulates such surveillance.
According to Apar Gupta of the Internet Freedom Foundation, the Delhi High Court had earlier allowed the use of facial recognition in the context of searching for missing children. But he argues that that basis cannot automatically be extended to monitoring peaceful protesters.
He cites the Supreme Court's KS Puttaswamy vs. Union of India decision, in which privacy was considered a fundamental right.
His question is that merely being present in a public place does not end a person's right to privacy and the state needs a clear legal basis to interfere in a person's privacy.
However, it is also important to note that the courts have not yet given any final verdict on the legality of the use of facial recognition at Jantar Mantar.
Delhi High Court has sought answers from the Center and Delhi Police in matters related to Jantar Mantar.
Apart from this, CPI(M) MP AA Rahim has also challenged the use of facial recognition and biometric surveillance in the Supreme Court.
Delhi Police, in an affidavit filed in the Supreme Court, has defended the use of Facial Recognition Technology (FRT) during the recent protests at Jantar Mantar.
Police say they did not conduct mass surveillance but used it to identify people with prior criminal records.
We also tried to contact Delhi Police repeatedly. We wanted to know that-
If facial recognition technology was used in the July 20 protest, which software was used?
From which database were the faces matched?
How long are recorded photographs and facial data kept?
Was any protestor identified and any action taken against him?
And what legal basis or SOP exists for this entire system?
After all, whose question is it?
In which database is the face being searched?
How long are photos kept?
Who will be responsible if the system misidentifies a person?
And if an algorithm deemed a person a suspect, how would he challenge that decision?
Perhaps the biggest question about AI surveillance is not how well it can identify us.
The question is also,
Are we able to recognize the system that is watching us?
