BOOM

Trending Searches

    SUPPORT
    BOOM

    Trending News

      • Fact Check 
        • Politics
        • Business
        • Entertainment
        • Social
        • Sports
        • World
      • ScamCheck
      • Explainers
      • News 
        • All News
      • Decode 
        • Investigations
        • Scamcheck
        • Features
        • Interviews
      • Media Buddhi 
        • Digital Buddhi
        • Senior Citizens
        • Resources
      • Web Stories
      • BOOM Research
      • BOOM Labs
      • Deepfake Tracker
      • Videos 
        • Facts Neeti
      • Home-icon
        Home
      • About Us-icon
        About Us
      • Authors-icon
        Authors
      • Team-icon
        Team
      • Careers-icon
        Careers
      • Internship-icon
        Internship
      • Contact Us-icon
        Contact Us
      • Methodology-icon
        Methodology
      • Correction Policy-icon
        Correction Policy
      • Non-Partnership Policy-icon
        Non-Partnership Policy
      • Cookie Policy-icon
        Cookie Policy
      • Grievance Redressal-icon
        Grievance Redressal
      • Republishing Guidelines-icon
        Republishing Guidelines
      • Fact Check-icon
        Fact Check
        Politics
        Business
        Entertainment
        Social
        Sports
        World
      • ScamCheck-icon
        ScamCheck
      • Explainers-icon
        Explainers
      • News-icon
        News
        All News
      • Decode-icon
        Decode
        Investigations
        Scamcheck
        Features
        Interviews
      • Media Buddhi-icon
        Media Buddhi
        Digital Buddhi
        Senior Citizens
        Resources
      • Web Stories-icon
        Web Stories
      • BOOM Research-icon
        BOOM Research
      • BOOM Labs-icon
        BOOM Labs
      • Deepfake Tracker-icon
        Deepfake Tracker
      • Videos-icon
        Videos
        Facts Neeti
      Trending Tags
      TRENDING
      • #Lok Sabha
      • #Narendra Modi
      • #Rahul Gandhi
      • #WhatsApp
      • #West Bengal
      • #BJP
      • #Deepfake
      • #Artificial Intelligence
      • #Scamcheck
      • Home
      • Decode
      • India’s Facial Recognition Drive On...
      Decode

      India’s Facial Recognition Drive On Hungry Children Is Erasing Them

      A Decode investigation finds that India’s use of facial recognition tech in its 50-year-old nutrition welfare programme is now extending to young children, creating new layers of exclusion.

      By -  Hera Rizwan |
      25 March 2026 12:17 PM IST
    • Boomlive
      Listen to this Article
      India’s Facial Recognition Drive On Hungry Children Is Erasing Them

      This series was supported by the Pulitzer Center.

      Ruhi (name changed) is four years old. Every morning, she walks to the government preschool in her village in Jharkhand, sits with the other children, and waits for her meal. According to the Poshan Tracker, the app that now controls India's child nutrition programme, Ruhi doesn't exist.

      She was born in a private hospital. The birth certificate her family received isn't recognised by the government. Getting an official one would require affidavits, witnesses, months of bureaucratic navigation, and fees for obtaining these documents, which her daily-wage labourer parents cannot afford. Without that certificate, she cannot get an Aadhaar biometric ID. Without Aadhaar, she cannot be enrolled for food.

      "Twelve children were removed from the app," said Shyamla Devi, who has run the Anganwadi centre in Chandankyari village for years. "But how do you tell a child not to come to the centre meant for children?"

      This is the second wave of exclusion in India's Integrated Child Development Services (ICDS)—a 50-year-old programme that feeds over 4.73 crore children, pregnant women and nursing mothers.

      In the first part of this investigation, Decode found pregnant women and nursing mothers are being denied food because facial verification technology cannot match their changing faces to years-old identity photographs. Now, that same system has been extended to children aged three to six.

      Also Read:AI Facial Recognition Is Denying Food To Pregnant Women Across India

      For children who do not yet have Aadhaar numbers, the system uses a parent's identity, usually the mother's. When facial verification fails to match the mother's current face to her database photograph, her child is dropped from the welfare rolls.

      Children now face a double trap: excluded either because they lack bureaucratic paperwork like Ruhi, or because algorithms cannot recognise their mothers' changing faces.

      A Decode investigation across Bihar, Jharkhand and Karnataka found children cut off from meals, preschool education and health monitoring because they lack paperwork, or because the facial recognition system cannot recognise their mothers' faces. Government data show that only 52.7% of eligible beneficiaries received rations by the end of 2025. The Ministry of Women and Child Development has not disclosed how many of those excluded are children.

      Shyamla Devi was instructed to remove at least four children from her centre's rolls after facial verification repeatedly failed to match their mothers' faces to database photographs.

      "They still come for morning snacks and to study," she said.

      "I can't turn them away just because the app didn't recognise their mother's face. I adjust whatever little allocation we get and feed them anyway."

      The Paperwork Trap

      Ruhi's exclusion did not begin with facial recognition. It began with a birth certificate.

      She could have been enrolled through her mother's Aadhaar, a practice that was common until recently. But her mother's Aadhaar photograph, taken years ago, no longer matches her current appearance after visible changes following childbirth. The AI-powered facial verification system fails to authenticate her.

      "The app doesn't recognise me," her mother said. "So we tried to make Ruhi's Aadhaar. But without a birth certificate, we can't."

      This double barrier—algorithmic rejection of mothers, bureaucratic rejection of children—is pushing families out entirely from the ICDS Scheme.

      Anganwadi workers across three states reported growing pressure to enrol children with individual Aadhaar numbers, despite policy and judicial limits.

      Under official rules of the Integrated Nutrition Support Programme, Saksham Anganwadi and Poshan 2.0 Rules, 2022—Aadhaar is not mandatory for children. If a child does not have Aadhaar, benefits must be delivered using a parent or guardian's Aadhaar identification. Yet children under five are being pushed toward Baal Aadhaar, which captures only a photograph and basic details. Officially voluntary, it has become the price of a meal.

      Delete Edit

      Neelam's Anganwadi centre in Karpi, Bihar has over 40 children enrolled in the 3-6 years-old category. (Image Credit: Tej Bahadur Singh)

      "Each child is now being treated as a separate data point," said one Anganwadi worker in Bihar who asked not to be named. "The system wants documents, linked mobile numbers, and verifications that families often lack."

      For years, the Anganwadi centre was Ruhi's family's safety net—nine people under one roof, dependent on daily-wage work when they could find it. Even when work was uncertain, one meal was assured. That certainty is gone.

      Hunger By Arithmetic

      Shyamla Devi feeds Ruhi anyway. She stretches the allocation meant for 30 enrolled children to cover 42 who actually come.

      The mathematics of this defiance are punishing. At an Anganwadi centre in Bihar's Karpi village, eggs are budgeted at six rupees though they cost eight. Oil is reimbursed at 86 rupees per litre; markets charge 150. Ghee is valued at 440 rupees; the real price is closer to 800.

      After those calculations, less than one rupee remains for fruit per child daily.


      "If we follow the rates exactly, there is nothing left before the month even ends," said Neelam, an Anganwadi worker since 2008.

      Centres receive vouchers, not cash, redeemable only at government ration shops with fixed prices. Workers cut eggs in half. Fruit becomes occasional. Some pay out of pocket until they can't.

      Vaishnavi Mangal, a doctoral researcher at the Norwegian University of Science and Technology who studies mobile phone use among frontline healthcare workers in India, frames the issue as one of skewed investment priorities.

      "The main purpose of an Anganwadi is simple," Mangal said. "The centres should function properly, services should reach people on time, and frontline workers should be able to do their jobs."

      Instead, she argued, public spending has tilted heavily toward digital oversight.

      "Technology cannot compensate for weak infrastructure. We are putting more money into tracking and reporting than into the nutrition and care the scheme is actually meant to provide."


      When Algorithms Meet Motherhood

      When the technology fails, a child is dropped from the rolls, cut off from food, preschool education, health monitoring and daily meals.

      The gatekeeper is a facial verification system embedded in the Poshan Tracker app, which uses multiple technologies including Google's ML Kit, a tool designed for tasks like unlocking smartphones. The AI tool is typically used in apps for tasks like photo filters and tagging faces. The faces it detects are then passed to an undisclosed system that performs the matching, ultimately determining whether children receive nutrition.

      Before any face is scanned, families must link Aadhaar to an active mobile number for one-time password verification. In households where phones are shared, lost, or travel with migrant workers, this step often fails.

      For those who reach the facial verification stage, the system compares live images against Aadhaar photographs often years old, poorly lit, and low-resolution. Faces changed by pregnancy, illness, manual labour, or aging fail to match.

      Responding to Decode's queries, Google said ML Kit itself is a general-purpose AI developer tool and not designed to identify specific people. The company said the software runs directly on the device and is provided as a freely available API used by developers around the world, meaning Google does not have visibility into how individual applications deploy it.

      “The Google ML Kit does not have facial recognition capabilities, and cannot identify specific people,” the company said, adding that developers who use the tools are responsible for the design, function and legal compliance of their applications.

      A Decode survey of 163 Anganwadi workers found three-quarters reporting frequent network failures during facial recognition. Each attempt takes over three minutes. Many workers spend three additional hours daily managing the system. Two-thirds said it made their work harder; a similar proportion fear delayed payments when the app fails.

      Decode surveyed Anganwadi workers across Bihar, Jharkhand and Karnataka on how facial recognition has changed their work

      A History Of Failed Fixes

      India has tried digitising ICDS before.

      In 2016, the Centre rolled out ICDS–CAS app, built by Delhi-based Dimagi Software Innovations with support from the Bill & Melinda Gates Foundation and a World Bank loan. Between 2017 and 2020, it covered about 6.4 lakh Anganwadis. But a 2022 BMJ Global Health evaluation found that while the system improved reporting and some service delivery indicators, it did not translate into measurable improvements in nutrition outcomes.

      In September 2020, ICDS–CAS was abruptly withdrawn with its data rendered inaccessible. Reports by NITI Aayog and the World Bank later cited network problems and duplication of paper and digital records, issues that preceded the rollout of the Poshan Tracker.

      The new system is owned by the Ministry of Women & Child Development and technically implemented by the National e-Governance Division (NeGD) under the Digital India Corporation, Ministry of Electronics and IT. It has since been scaled nationally and linked directly to Anganwadi workers’ payments.

      On the ground, however, the same problems persist. The addition of AI-powered facial verification has raised the stakes, turning routine verification into a fragile checkpoint that can exclude children and deepen stress for Anganwadi workers. The platform has changed, but the structural obstacles remain.

      Frequent app crashes force Shyamla to log data in her register. (Image Credit: Tej Bahadur Singh)

      According to researcher Vaishnavi Mangal, the issues evident under the ICDS–CAS app have persisted, if not intensified, with the rollout of the Poshan Tracker. “The problems are continuing; phones malfunction, internet lags, and the system keeps breaking down,” she said. While ICDS–CAS had provisions for offline data entry, she pointed out, Poshan Tracker depends heavily on live connectivity, making it far less usable in low-resource settings.

      Because of this, Anganwadi workers do not fully trust the app, Mangal explained, and continue to maintain physical registers alongside it.

      “They are not data entry operators, they are caregivers. But most of their time is now going into uploading data instead of actually doing care work,” she said.

      In her assessment, Poshan Tracker has effectively become a parallel reporting system built for officials at the top. “It does not seem to be benefiting Anganwadi workers or the beneficiaries,” she added.

      Where Children’s Data Goes

      For families visiting Anganwadi centres, it often feels simple. A child is weighed, a photo is taken, a meal is served. What happens next is invisible.

      Poshan Tracker runs on Amazon Web Services. India's largest child nutrition database—biometric data on millions of children—sits on infrastructure owned by a foreign corporation. The government claims ownership, but processing and storage could be happening outside its direct control. When the system was scaled after 2021, the National e-Governance Division partnered with Daffodil Software, a private IT firm, to build the platform.


      The Poshan Tracker holds names, Aadhaar numbers, pregnancy status, delivery dates, nutrition records, health indicators, and location data. The government says photos aren't permanently stored on workers' phones, and temporary images are deleted at logout. But no public information exists on retention periods, access controls, data residency configurations, or whether images are reused.

      The Ministry of Women and Child Development did not respond to Decode’s RTI queries seeking clarity on the handling of sensitive data within the app, or on whether the system has undergone any third-party evaluations or audits.

      Decode’s investigation found that the system is closed and unauditable. This means that there is no way for researchers to independently assess how securely this sensitive data is stored, or how prepared the government is to prevent or respond to a breach.

      Advocate Alvin Antony, Chief Compliance Officer at GovernAI, said that when a public welfare system relies on foreign-owned cloud infrastructure, the government’s control over data is no longer purely sovereign.

      “The State’s de facto control is mediated through contracts rather than public law,” he explained.

      In practical terms, this means that if there is a data breach, a technical failure, or a foreign legal order, beneficiaries may find it difficult to seek remedies because “the key evidence and technical capability lie with the cloud provider, not directly with the Indian government”.

      Antony noted that this imbalance matters even more because the data involved belongs to children and low-income women, groups that have limited ability to litigate or exercise data rights.

      The issue becomes more complex under the US CLOUD Act, which allows US authorities to demand data from US-based companies even if that data is stored outside the country.

      Antony pointed out that India has deliberately chosen not to sign a CLOUD Act agreement with the US, which helps keep such access “at arm’s length”. However, he added that US authorities can still seek data under their domestic law, though this requires judicial warrants and is subject to strict legal standards.

      On the Indian side, Antony said the main safeguards come from the Digital Personal Data Protection Act, 2023, and its rules, which mandate security measures, limited data retention, and penalties for breaches. But these protections, he cautioned, are largely procedural. “They regulate how data is processed and secured, not whether people— or even Indian authorities— will be informed when a government seeks access.”

      The result, he explained, is a system where beneficiaries have little visibility into “where their data goes, who controls it, and what remedies exist if something goes wrong,” even though the information being collected is deeply sensitive and tied to basic welfare entitlements.

      Consent As Digital Fiction

      The app carries a declaration: the worker has informed the beneficiary their sensitive data will be used by the government. In practice, workers pressed for time and fighting app crashes tap through without explaining. For families with limited literacy, consent is a box the system checks, not a choice they make.

      India's Digital Personal Data Protection Act (2023) promises safeguards for children's data including parental consent. But it grants broad exemptions for government welfare delivery. Rules were notified recently with an 18-month rollout. Poshan Tracker, built earlier, operates in this gap.

      Alluding to India's data protection law, Antony explained that while the Act technically applies to systems like Poshan Tracker, its protections for children are weak in practice.

      "On paper, the law gives children special safeguards. But because the State can process data for welfare without consent, those safeguards are easily bypassed," he said.

      In schemes like these, Antony said, parental consent becomes largely meaningless because biometric authentication is mandatory to receive benefits. "If refusing facial recognition means losing food or care, that's not real consent."

      He also pointed out that the law gives the government wide exemption powers. "The State can exempt itself from many obligations in the name of public interest. So while the law talks about children's rights, government-run welfare systems can legally operate with very limited accountability."

      The result, he said, is a gap between promise and practice: "Children's privacy is recognised in theory, but the structure of the law allows it to be diluted when the government is the data collector."

      Delete Edit

      Height and weight data of children are routinely logged on the app (Image Credit: Tej Bahadur Singh)

      The gap between law and practice extends beyond data protection. Advocate Dipika Sahani pointed out that the pressure to enrol children with individual Aadhaar directly conflicts with the Supreme Court's 2018 ruling, which was unequivocal that welfare benefits cannot be denied due to authentication failures and that beneficiaries must always have alternative ways to establish identity.

      "Any privacy infringement has to be proportionate," Sahani said, adding that making food access contingent on biometric verification effectively coerces consent "under the threat of starvation".

      The Bombay High Court went further. In rulings in 2018 and 2021, it ordered foodgrains to be distributed irrespective of Aadhaar-linking issues, clearly stating that "technology shall not be a rider for implementation of the scheme" and that technical lapses cannot override a fundamental right.

      Yet on the ground, children like Ruhi are being excluded for lack of Aadhaar, and others are being removed when their mothers fail facial recognition—direct violations of these judicial pronouncements.

      Exporting The Experiment

      What is unfolding inside India’s welfare infrastructure is not confined to its borders. Following India’s G20 presidency, a broader model of tech-enabled governance—often framed as efficient, scalable, and inclusive—is now being promoted abroad. India’s Digital Public Infrastructure, including Aadhaar, Unified Payments Interface and DigiLocker, is being taken to countries across Africa and Asia, with support from institutions such as the World Bank and the United Nations.

      Mila Samdub, an independent expert on the political economy of digital infrastructure, cautioned that the risks travel with the technology. When such systems are exported, he said, they often fail if they are not designed around the needs of end users or do not translate into real improvements in people's lives.

      "The key question is what happens when these systems fail," he said.

      With Aadhaar coverage still low, only a few children undergo face scans. (Image Credit: Tej Bahadur Singh)

      In some cases, Samdub explained, technology-led welfare projects simply collapse after a few years, leaving behind wasted public money and political capital. This, he said, has been a familiar pattern in many "technology for development" interventions.

      More troubling, however, is the alternative scenario. “These systems may be pushed through regardless of their flaws,” he warned, forcing people to “adjust their lives to the requirements of the system,” often at the cost of dignity and access.

      Underlying this, Samdub pointed to a deeper design assumption in India's digital governance model: that more data automatically leads to better outcomes.

      "Consent frameworks under India's digital protection law are structured to maximise data collection and flow," he said.

      What is missing, he argued, is independent and rigorous evidence of impact.

      The Workers Who Feed Ghosts

      For Anganwadi workers, the consequences of these design choices are felt in everyday work—serving meals, marking attendance, and keeping records.

      The app also functions as a monitoring tool for workers themselves. Anganwadi workers told Decode they are required to open centres 22 of 25 days monthly, verified through the app. Network failures require workers to travel to block offices with written explanations.

      Anganwadi workers earn an honorarium between Rs 9,000 and Rs 15,000 a month, payments that are often delayed and, in many states, at or below the notified minimum wage for an eight-hour workday. Yet many say they routinely work nearly 10 hours daily, balancing preschool teaching, nutrition distribution, home visits, paperwork, and now mandatory digital uploads. Their pay increasingly hinges on whether photos and entries are successfully uploaded on an app that frequently malfunctions.

      “Earlier, if something went wrong, we could explain it,” said an Anganwadi worker from Jharkhand.

      “Now the app decides whether we worked or not, leaving us with no agency. Even if the food is cooked and the children have eaten, if the photo or entry doesn’t upload, it is treated as if nothing happened.”

      Many workers describe their situation as "like forced labour". Most are middle-aged women who have spent years in this role, with few alternative livelihoods. Leaving is not a real option; staying means complying with a system that constantly threatens their income.

      This comparison is not just rhetorical.

      Nirmal Gorona, convenor of the National Campaign Committee for the Eradication of Bonded Labour, explained that Indian courts have long recognised economic compulsion as a form of force. In the Asiad Games case, the Supreme Court held that when people work for less than the minimum wage because they have no real alternative, it can amount to forced labour under Article 23.

      That principle was later applied by the Allahabad High Court to mid-day meal cooks, who were paid honoraria far below minimum wages. “What we are seeing with Anganwadi workers fits the same pattern as they are putting in 10–12 hours a day for less than minimum wages,” Gorona said.

      When workers get stuck in the app, support is minimal. Poshan Tracker includes a grievance redressal mechanism through a ticketing system, but most workers Decode spoke to said they were unaware of it. The Ministry also runs a toll-free helpline 1515, earlier 14408, managed by the National e-Governance Division under the Digital India Corporation.

      Officially, complaints are handled by call-centre executives or forwarded to concerned authorities. However, a reply to an RTI query by Decode, showed that complaints received through the helpline and ticketing system are not segregated between Anganwadi workers and beneficiaries, leaving no public data on how many grievances are addressed.

      Delete Edit

      Anganwadi workers stretch rations to feed children left out of the system. (Image Credit: Tej Bahadur Singh)

      Ruhi still walks to the Anganwadi centre every morning, hoping for an egg at breakfast and rice porridge at lunch. The app, however, continues to say she does not exist.

      Her mother, anxious about what lies ahead for the family, says,

      “Ruhi has a younger brother too. The same thing will happen to him. We never knew that choosing a private hospital for childbirth would decide so much about our children’s future.”

      Across India, where workers cannot stretch allocations further or where the app's logic is enforced more strictly, children are disappearing not from hunger, but from the system meant to prevent it.

      This story has been edited by Adrija Bose.

      Tags

      Anganwadi WorkersMinistry of Women and Child Developmentbiometric technologyfacial recognitionArtificial Intelligence
      Read Full Article

      Next Story
      X

      Subscribe to BOOM Newsletters

      👉 No spam, no paywall — but verified insights.

      Please enter a Email Address
      Subscribe for free!

      Stay Ahead of Misinformation!

      Please enter a Email Address
      Subscribe Now🛡️ 100% Privacy Protected | No Spam, Just Facts
      By subscribing, you agree with the Terms & conditions and Privacy Policy connected to the offer

      Thank you for subscribing!

      You’re now part of the BOOM community.

      Or, Subscribe to receive latest news via email
      Subscribed Successfully...
      Copy HTMLHTML is copied!
      There's no data to copy!