{
  "type": "article",
  "title": "How Facial Recognition Malsidentified 25 Incarcerated Individuals During Jantar Mantar Protest",
  "summary": "An investigation revealed that at least 25 individuals who were locked up in jail at the time were flagged by Delhi Police facial recognition cameras during a protest at Jantar Mantar.",
  "content": "The Supreme Court recently quashed all FIRs related to the exam-leak protests while allowing the government to proceed against 2,873 individuals flagged by Delhi Police at Jantar Mantar. Police stated that these individuals possessed criminal records and were identified through facial recognition software between 20 and 26 July. Authorities also emphasized that no punitive measures would be executed based solely on facial recognition, noting that every single case must undergo prior field verification.\n\nAn investigation uncovered that at least 25 people were actually behind bars when the system identified them at the demonstration. These individuals included people facing serious charges such as murder, attempted murder, rape, and child abuse. The Cockroach Janta Party staged a protest at Jantar Mantar on 6 June 2026 to demand fairness regarding examination issues and the resignation of then Education Minister Dharmendra Pradhan. A march toward Parliament was subsequently called for 20 July, with the controversial facial recognition scans taking place between 20 and 26 July.\n\nDelhi Police informed the Supreme Court that its Automated Facial Recognition System identified 2,873 people with criminal backgrounds, out of which 2,402 were detected through Crime Kundli and 471 through alternative criminal records. On 1 September, the Supreme Court quashed the FIRs connected to the protests while allowing proceedings against the 2,873 flagged names to continue. However, this inclusion does not automatically classify all 2,873 individuals as convicted criminals, as the police affidavit failed to explicitly establish the distinct legal status of every single person on the list.\n\n \n\nWhy Delhi Police Deployed Facial Recognition on the Crowd\n\nDelhi Police utilized Automated Facial Recognition Systems to spot individuals with prior criminal antecedents directly within the protesting crowd. The technology allowed law enforcement to compare faces captured from live video feeds against photographs already stored inside police databases. Authorities argued that the deployment aimed to pinpoint individuals with severe criminal histories to assist public safety and crowd management during demonstrations. Delhi Police maintained in court that facial recognition outputs serve merely as initial investigative leads, requiring rigorous human verification before any legal action is enacted. The system was never designed to independently decree someone a criminal, but rather to flag potential matches for human officers to investigate.\n\n \n\nWhere the System Retrieved Matching Photographs\n\nThe police software cross-referenced faces captured within the crowd against photographs and historical files residing within official police archives. A primary database cited by Delhi Police was Crime Kundli alongside other specialized criminal record repositories. These archives consist of photographs and personal details gathered by law enforcement during previous investigative proceedings and criminal trials. Consequently, the technology was not identifying individuals from scratch, but rather matching newly captured protest faces against preexisting photographs inside official databases. According to Pavan Duggal, a cyber law expert and advocate practicing in the Supreme Court of India, government utilization of AI surveillance systems without explicit parliamentary authorization is legally flawed, highlighting an urgent need for regulatory frameworks governing AI accountability.\n\n \n\nHow Incarcerated Individuals Were Flagged at the Protest\n\nFacial recognition software possesses no inherent awareness of a person's physical location in the real world. The system analyzes a face, generates a digital mathematical template, and contrasts it against stored database portraits. When the software encounters a sufficiently high similarity score, it registers a potential match. Key variables affecting accuracy include the quality of database photographs and the system matching threshold established by operators. Delhi Police previously noted that its system treats any match reaching an 80 percent threshold as positive. Nevertheless, an 80 percent threshold does not equate to an 80 percent accuracy rate or a 20 percent error rate. Available evidence confirms these incarcerated individuals were flagged despite being imprisoned, but it does not pinpoint the exact technical breakdown in each individual instance.\n\n \n\nError Rates and Limitations of Facial Recognition\n\nFacial recognition technology features no single universal error rate, as performance fluctuates based on software algorithms, camera hardware, image resolution, database composition, and matching thresholds. Extensive research conducted by NIST, MIT, and Stanford indicates that facial recognition systems generate both false positives and false negatives regularly. Studies demonstrate that error rates vary considerably across different algorithms and demographic groups. Within a massive crowd, even a modest error rate can trigger numerous false matches due to the sheer volume of faces compared against extensive databases. This inherent volatility underscores why facial recognition results must be treated solely as investigative leads requiring strict human verification rather than conclusive evidence.\n\n \n\nData Retention and Privacy Concerns\n\nIndia lacks a unified nationwide retention period specifically regulating every police deployment of facial recognition technology. While the Digital Personal Data Protection Act of 2023 exists, it does not establish a dedicated, comprehensive regime governing data retention timelines for police facial recognition systems, while containing broad exemptions for state functions. Consequently, retention durations depend entirely on the specific agency, platform, and legal framework involved. Authorized government agencies maintain access to matching databases, raising profound privacy concerns regarding cross-agency data sharing where information gathered for one purpose gets repurposed elsewhere. The ultimate fate of facial recognition data collected during public protests remains one of India's most pressing unanswered privacy questions.\n\n \n\nLegal Standing of Facial Recognition Matches\n\nA facial recognition match alone does not constitute legal proof that an individual attended a protest or perpetrated a crime. Delhi Police formally assured the Supreme Court that no actions are taken solely based upon facial recognition outcomes. According to official statements, the software merely generates potential leads, obligating officers to verify identities and corroborate other evidence before moving forward. This distinction is critical regarding the Jantar Mantar incident, where flagging individuals who were physically imprisoned at the time demonstrates that police must cross-check custody records before treating algorithmic matches as reliable. Therefore, facial recognition offers preliminary leads but cannot independently establish physical presence or criminal culpability.\n\n \n\nConsequences of Algorithmic Misidentification\n\nA false positive match carries severe real-world consequences, even when subsequent corrections clear the accused individual. Citizens face wrongful police interrogations and investigations simply because flawed algorithms link their facial features to someone else's criminal file. Beyond misidentifications, profound questions persist regarding surveillance oversight, civil liberties, and due process. Facial recognition extends far beyond a mere technological tool, involving fundamental questions of privacy, accountability, and human verification. For instance, in the United States, Nijeer Parks was wrongfully arrested in 2019 after police acted on an unreliable facial recognition match, illustrating the urgent need to safeguard fundamental constitutional liberties against automated overreach.\n\nWhat this means for you\nThis malfunction in facial recognition technology carries profound practical implications for civil liberties, citizen privacy, and police surveillance transparency across the nation.\n\n• Across India: Ordinary citizens participating in public demonstrations face the alarming risk of algorithmic misidentification and wrongful police flagging. This threatens fundamental democratic rights regarding peaceful assembly and public expression.\n• In Delhi: Local residents and activists risk undergoing unwarranted police interrogation and harassment due to automated false positives without rigorous human verification. Individuals may need to navigate legal safeguards to protect personal liberty against automated errors.\n\nQuestions & Answers\n\n1. When was the protest held at Jantar Mantar?\nThe Cockroach Janta Party held a protest at Jantar Mantar on 6 June 2026 regarding examination issues.\n\n2. How many individuals were flagged by Delhi Police?\nThe facial recognition system identified 2,873 people with criminal antecedents.\n\n3. How many flagged individuals were actually in jail?\nAn investigation found that at least 25 people were locked up in prison when the system flagged them.\n\n4. What criminal charges were the incarcerated individuals facing?\nThey included individuals facing charges of murder, attempted murder, rape, and child abuse.\n\n5. Which database was primarily used for matching faces?\nPolice cross-referenced faces primarily using Crime Kundli along with other official criminal databases.\n\n6. Is a facial recognition match treated as conclusive legal proof?\nNo, police stated that matches serve only as initial leads requiring mandatory human verification.\n\n7. What did the Supreme Court decide regarding the protest FIRs?\nOn 1 September, the Supreme Court quashed the FIRs related to the exam-leak protests.\n\n8. What did legal expert Pavan Duggal say about the system?\nPavan Duggal stated that using AI surveillance without parliamentary authorization is wrong and requires accountability laws.",
  "url": "https://trendkia.com/en/investigations/how-facial-recognition-malsidentified-25-incarcerated-individuals-during-jantar-mantar-protest-27942",
  "category": "Investigations",
  "publishedAt": "2026-09-05",
  "tags": [
    "facial recognition",
    "Delhi Police",
    "Jantar Mantar protest",
    "surveillance technology",
    "Supreme Court",
    "Crime Kundli",
    "AI surveillance"
  ],
  "language": "en",
  "site": "TrendKia"
}