Jailed accused among those ‘identified’ at Jantar Mantar protest

An investigation into the Delhi Police’s use of facial recognition technology during the NEET-UG protests has found that at least 25 people identified by the system at Jantar Mantar were in prison at the time, raising questions about the accuracy of the technology and its use to monitor peaceful assemblies.

The finding assumes significance as the Supreme Court examines whether the deployment of the Automated Face Recognition System (AFRS) and other biometric tools during the protests had a legal basis and whether such surveillance violated the fundamental rights to privacy, free speech and peaceful assembly.

According to an investigative report published by the Indian Express, the 25 people were among 2,873 individuals whom the Delhi Police claimed had criminal antecedents and were identified by facial recognition software at the main protest venue between 20 and 26 July.

However, police, prison and court records examined as part of the investigation showed that the 25 individuals were lodged in Delhi’s Tihar, Mandoli and Rohini prison complexes when the system purportedly detected them at Jantar Mantar.

The findings came days after the Supreme Court quashed FIRs registered across the country in connection with the student protests. The court, however, allowed the authorities to proceed against the 2,873 people whom the Delhi Police said had criminal records and had been identified at the protest.

The Delhi Police has stressed that facial recognition results are not treated as conclusive proof of a person’s presence and that no action will be taken without field verification.

Responding to questions from the Indian Express, the police said: “During the protests, 2873 individuals identified as prima facie having criminal antecedents were spotted. However, further verification in this regard is pending.”

Murder, rape accused among prisoners flagged

The media investigation examined 205 people from the police list who had been booked for particularly serious offences. The break-up included 101 people accused of murder, 61 of rape, six booked under the Protection of Children from Sexual Offences Act and 37 of the 62 people listed under attempt to murder.

Records showed that 17 people facing murder charges, four linked to rape cases and four accused of attempt to murder were in prison when the facial recognition system allegedly identified them at the protest.

Of these 25 individuals, three were shown as having been identified on 24 July, 21 on 25 July and one on 26 July, the final day of the Jantar Mantar protest.

Those facing murder charges and shown as being in prison included Kishan, Mustaq, Ankit Kumar, Yogesh, Pratap Singh Sisodia, Arjun, Sonu Kumar, Vijay, Rakesh, Bhola alias Shyam Kumar, Raja, Vicky alias Manoj Thakur, Mohd Faraz, Manish, Dharmender Yadav, Sonu and Royal Massy.

Among them was Yogesh alias Raju, who was arrested for his alleged involvement in the killing of gym owner Nadir Shah in Greater Kailash-1 in south Delhi. Police records describe him as a sharpshooter associated with the syndicates of Lawrence Bishnoi and Hashim Baba. He had been arrested following an exchange of fire in which he suffered a bullet injury.

The four people linked to rape cases who were reportedly in prison were Ankit alias Mogli, Sunil Kumar, Heera Singh and Kundan. The latter three had been arrested in cases registered under the POCSO Act.

Mahesh, Bobby, Jishan Ahmad Idrishi and Annu Gilodia alias Pahalwal, all linked to attempted-murder cases, were also shown by prison records to have been behind bars when the facial recognition system flagged them at the protest.

Police affidavit cited 2,873 matches

In an affidavit submitted to the Supreme Court on 17 August, the Delhi Police said the system had identified people whose particulars matched records available in official police databases.

Of the 2,873 matches, 2,402 were reportedly generated through the Delhi Police’s biometric database, Crime Kundli, while 471 came from criminal records. The police said the list included 92 people allegedly involved in more than 10 cases and 47 described as history-sheeters. The offences associated with those identified included murder, rape, drug-related crimes and cases under the POCSO Act.

The affidavit, however, did not specify whether every person identified was an accused, a convict or merely named in a criminal case.

According to the police, its records contain facial images and other information relating to people accused of serious crimes and not those involved in minor infractions such as traffic violations.

The Delhi Police further said that the software did not automatically create profiles of peaceful protesters and was used only to flag people with previous records involving serious offences.

“No actions are taken solely based upon the result of the facial recognition system which is only the first step to identify a person in the crowd who has a criminal record,” the affidavit said.

It added that field verification was conducted after a face was recognised and that action was contemplated only if the subsequent inquiry established that the person had actually been present at the protest.

The apparent identification of people who were in prison, however, places the reliability of those preliminary matches under scrutiny.

Guruswamy questions legal basis of surveillance

The constitutional challenge to the use of AFRS has arisen from a public interest litigation petition filed by CPI(M) Rajya Sabha MP A.A. Rahim. Senior Advocate Menaka Guruswamy and Advocate Subhash Chandran are representing Rahim in the Supreme Court.

The petition has been tagged with other pleas concerning alleged police action during the NEET-UG protests and demonstrations associated with the “Cockroach Janta Party” campaign.

It alleges that thousands of protesters, journalists and ordinary citizens were subjected to “continuous and pervasive biometric surveillance” through CCTV cameras, drones, handheld devices and a Mobile Command and Control Vehicle.

The petition also refers to the use of the Ikshana surveillance vehicle, facial recognition-enabled AjnaLens smart spectacles and the collection of fingerprints through the National Crime Records Bureau’s Abhigyan application, which is connected to the National Automated Fingerprint Identification System.

The plea alleges that biometric information was processed and hosted by private companies, including Aditya Infotech Ltd and Dimension NXG Pvt Ltd, without clear statutory authority or adequate safeguards for the collection, storage, use and deletion of such data.

During the Supreme Court proceedings, Guruswamy said: “One maps your face. Spectacles are used and a vehicle is also used. The data is taken without permission. Then private entities host the data in violation of DPDP Rules, CrPC and a variety of things.”

Emphasising that the challenge concerned the use of facial recognition against protesters, she told the court: “This is very specific prayer concerning facial recognition being used by the Delhi Police against protesters, and the data is held with private companies.”

Guruswamy also argued that private companies were being used to process the information despite the absence of a dedicated regulatory framework. Responding to the government’s claim that the technology captured only people accused of serious offences, she said: “They collect all pictures.”

The petitioners contend that a facial recognition system must first capture and process the faces of everyone appearing in a video frame before it can compare them with a watchlist. Therefore, they argue, the initial collection itself amounts to indiscriminate biometric surveillance, irrespective of whether the software eventually flags only those with criminal records.

Government says matches undergo verification

Solicitor General Tushar Mehta, appearing for the Union government and the Delhi Police, rejected suggestions that the system was being used indiscriminately against students and other peaceful protesters.

“Facial recognition only captures hardened criminals. They are people who are in [NCRB] data, accused of serious offenses. Action is not taken immediately. If face matches, then there would be field verification and investigation,” he told the court.

The authorities have also maintained that the videography at Jantar Mantar was intended to maintain law and order and ensure public safety, rather than to snoop on protesters.

The facial recognition software places boxes around faces detected in footage and compares their digital templates with images stored in police databases. If the similarity score crosses a predetermined threshold, the system produces a possible match for further examination.

In response to a Right to Information request in 2022, the Delhi Police disclosed that it considered a facial recognition result positive if the system showed an accuracy level of 80 per cent.

Such a result, however, does not by itself establish identity. The reliability of a match can be affected by the algorithm used, the quality and angle of the footage, lighting, masks and the database against which the image is compared.

Studies by the US National Institute of Standards and Technology have also found differences in error rates across demographic groups in several facial recognition algorithms, adding to demands for human verification, independent audits, transparent thresholds and detailed records of how matches are generated.

Vrinda Grover seeks gender-sensitive oversight

Senior Advocate Vrinda Grover, who appeared in the connected proceedings concerning police action against the student protesters, sought the quashing of FIRs registered against peaceful demonstrators across different states.

Grover also emphasised the need for gender-sensitive oversight while the court considered the creation of a high-powered committee to examine protest-related grievances. She argued that women officers should be included in any such panel because several of the issues before the court affected women protesters.

“We can recommend names of former women officers,” she told the court while discussing the composition of the proposed committee.

Grover also argued that counter-affidavits addressing allegations such as the use of pellet guns should be placed before the Supreme Court, saying: “If counter affidavit on issues like pellet guns are filed before court, it will actually assist committee.”

During the hearings, petitioners’ counsel also maintained that the constitutional validity of facial recognition surveillance could not be delegated to a committee and required a judicial determination.

“Since there are issues related to women, there should be a woman in the committee. I had also raised aspect of facial recognition. That relates to constitutionality. That can’t be sent to Committee,” the court was told on behalf of the petitioners.

SC to decide constitutional questions

A three-judge Bench comprising Chief Justice of India Surya Kant and Justices Joymalya Bagchi and V. Mohana has issued notice to the Centre and the Delhi Police Commissioner over the challenge to the surveillance.

The court has directed the authorities to preserve all data collected during the protests and ensure that it is not placed in the public domain.

It has indicated that a high-powered committee may examine several questions arising from the handling of the protests. At the same time, the Bench has clarified that the constitutional questions surrounding facial recognition technology will be determined by the Supreme Court itself.

The petitioners have relied on the Supreme Court’s 2017 K.S. Puttaswamy judgment, which recognised privacy as a fundamental right. Under the principles laid down in that verdict, an intrusion on privacy must have a legal basis, pursue a legitimate state objective and satisfy the test of proportionality.

They argue that no specific law authorises the indiscriminate biometric monitoring of people attending a peaceful assembly, that the surveillance is broader than required for maintaining public order and that there are inadequate safeguards governing the use and retention of the data.

The plea also invokes the Digital Personal Data Protection Act, 2023, alleging that the collection and processing of biometric data without consent and its handling by private companies violate principles of purpose limitation, accountability and data protection.

The Delhi Police may rely on Section 17 of the law, which provides exemptions from certain obligations when personal data is processed for preventing, detecting, investigating or prosecuting an offence. The Centre can also exempt notified state instrumentalities for reasons including public order and national security.

However, Indian law does not provide a dedicated framework governing the police use of facial recognition. There are no comprehensive statutory rules specifying how watchlists should be created, the accuracy threshold required for a match, how long protest footage may be retained or whether independent audits must be conducted.

The police have reportedly said that there are no fixed retention guidelines for the footage collected at the protest because it could be required for a future investigation.

The legal dispute therefore extends beyond whether individual matches were correct. The larger constitutional question is whether continuous identification of people attending a peaceful protest is a narrowly tailored public-safety measure or a disproportionate intrusion that treats every participant as a potential suspect.

The reported identification of 25 people who were already in prison is likely to sharpen that debate when the Supreme Court considers the accuracy, legality and safeguards surrounding the use of facial recognition in protest policing.

With agency inputs