STATE DEPLOYMENT OF ARTIFICIAL INTELLIGENCE AND THE RECONFIGURA-TION OF CONSTITUTIONAL SAFEGUARDS
“For Your Safety & Our Curiosity” – This ironic slogan captures the tension between India’s security narrative and individual privacy. In recent years the Indian government has woven complex data-driven technologies into governance: hundreds of millions of citizens are monitored or profiled through Aadhaar biometrics, CCTV networks, “social registry” databases, health apps, and welfare algorithms. Algorithmic governance – broadly, governance by data and automated rules – means that social ordering now “relies on coordination between actors, is based on rules and incorporates particularly complex computer-based epistemic procedures”1. In practice, these systems promise efficiency and transparency but also raise grave privacy questions. India’s Supreme Court has recognized that privacy is a constitutional right (K.S. Puttaswamy, 2017) and warned that advancing technology provides “fresh machinery for possible infringement of privacy by the State through surveillance, profiling and compilation of data”2. Yet, without strong legal guardrails, the state’s covert use of AI can quietly erode the informational privacy that the Constitution guarantees. This paper surveys current evidence of India’s AI deployments across the country and examines how they strain existing safeguards.
Digital transformation in India has been profound. Large-scale digital identity (Aadhaar), ubiquitous smartphones, and e-governance platforms have generated vast datasets and enabled algorithmic governance – the use of AI and automated decision-making (ADM) in public administration. Globally, governments are rushing to apply AI “to improve efficiency and responsiveness” in public services. A recent OECD study notes that “67% of OECD countries [are] using AI to improve” public service delivery, aiming to optimize resources, automate routine processes, and tailor services to citizens. In theory, AI can make governance faster and seemingly fairer (being “less prone to human error or corruption”). India too has embraced digital innovation: from the JAM trinity (Jan Dhan bank accounts, Aadhaar ID, Mobile connectivity) to a host of digital public goods (e.g. the Unified Payments Interface, NDHM health IDs, etc.). The government’s own AI guidelines emphasize inclusive development and accountability, mandating that AI be “designed and tested to ensure outcomes are fair, unbiased, and do not discriminate”. They stress that AI must be understandable by design, with “clear explanations” for users and regulators.
Yet alongside these promises lie serious concerns. In practice, algorithmic governance can erode fundamental rights and privacy in subtle ways. Systems are often opaque and proprietary, raising questions of transparency and due process. Bias in training data can embed historical prejudices into automated decisions, with potentially discriminatory effects. Errors – e.g. false matches in facial recognition – can lead to wrongful actions with grave human costs. In India’s legal context, constitutionally guaranteed equality (Article 14) and non-discrimination (Article 15) “provide a constitutional foundation for challenging discriminatory AI systems”. But current laws (the IT Act, even the new Data Protection Act, 2023) lack explicit AI safeguards. There is no dedicated regulator for AI justice, so citizens harmed by algorithms often face “significant barriers to justice”. Interdisciplinary scholars and technologists warn that without robust oversight, India could see a “quiet erosion” of privacy and autonomy under the guise of tech efficiency.
STATE SURVEILLANCE AND POLICING
India has rapidly expanded mass surveillance with AI tools. The government’s legacy systems (e.g. the Central Monitoring System and biometric Aadhaar) have been joined by smart cameras and algorithms that analyze citizens’ data in real time. For example, the GIGA Institute reports that AI-enabled facial-recognition cameras and drones “have been mainstreamed into public life without statutory basis or the consent of the surveilled”3. Delhi police alone retrofitted thousands of CCTV cameras with face-recognition to probe the February 2020 riots: official data show FRT was used in 750 riot-related cases and “presented as evidence” in 98.9% of them4. (Subsequent reports indicate that over 80% of those cases ended in acquittals5 underlining the technology’s high false-positive rate.) Today many states use such systems routinely – traffic cameras scan license plates, predictive-analytics tools sort criminal complaints, and AI is deployed in policing hot spots – often supplied by private firms but used by public agencies without transparency. In effect, a city or district can become a data “dragnet” where everyone is potentially a suspect.
These technologies shift police work onto algorithmic platforms. Even apart from riots, police forces in multiple cities have adopted AI tools: systems like “TriNetra” in Chennai and “ABHED” in Rajasthan scan CCTV feeds for wanted persons, while programs in Maharashtra and Telangana match mugshots against footage and social-media images. As per George & Sinha (2025) as one investigation notes, “police in several Indian states have routinely made use of fingerprint and facial recognition technology to stop and screen people… turning vital public spaces into privacy-violating zones”. All of this occurs with little legal restraint. There is no specific law governing police use of AI or required oversight, so decisions about who is watched, how data is stored, or how matches are verified lie in bureaucratic hands. In practice, citizens have no easy way to learn if an algorithm flagged them as suspicious, much less to challenge it. The cumulative effect is a quiet expansion of state gaze – cameras and algorithms collect and sift personal data continuously, even as formal privacy protections lag behind.
WELFARE ALGORITHMS AND SOCIAL DATABASES
Government AI isn’t confined to policing. In welfare administration, data-driven systems are emerging to determine eligibility and root out fraud – with alarming side-effects. Several states are consolidating disparate databases into “social registries” and running algorithms over them. For instance, Telangana’s Samagra Vedika platform illustrates how entity‑resolution systems can quietly reshape welfare governance. The system links records from multiple state databases, including welfare schemes, property and other registries, to construct unified profiles of individuals and households and flag “duplicate” or “fraudulent” beneficiaries for administrative action. Amnesty International’s technical assessment shows that these matching processes depend on opaque algorithms and imperfect data, creating a significant risk that eligible people will be mis‑identified and excluded from essential entitlements.6
Telangana’s Samagra Vedika system creates a “360-degree” digital profile of each citizen by merging data from income records, public distribution ledgers, property registries, and more7. Under state policy, village officials were required to defer to Samagra Vedika’s recommendations when issuing or cancelling food-security (ration) cards[9]. Initially pitched as a way to weed out bogus beneficiaries, the system ended up denying millions of the poor their rightful subsidies.
Tapasya & Doshi (2024) in their analytical report show that in one documented case, an algorithm erroneously tagged a rickshaw-puller as the owner of a car, causing his widowed wife to be cut off from subsidized food during the pandemic. It took months of appeals before officials overruled the computer. The report further goes on to showcase the investigations found that between 2014-19 Telangana cancelled about 1.86 million food entitlements and rejected 142,000 new applications, claiming these were fraudulent. However, journalists later showed that “several thousands” of those exclusions were wrongful – caused by bad data or “faulty data and bad algorithmic decisions” by the system. In effect, low-income families were forced to prove their poverty to machines rather than to humans. Similar efforts are underway in other states: social registries and Aadhaar-based eligibility checks are linked to pensions, housing, and health schemes. Nationally, reports surfaced in 2020 that the central government proposed a unified “360-degree database” of citizens (a Social Registry Information System) incorporating Aadhaar to track all aspects of life8.
At a broader level, automated welfare can reinforce India’s deep digital divides. Studies show that biometric or algorithmic processes often disadvantage the rural poor, women, and migrants. For example, a survey of migrant female workers found 36% had experienced Aadhaar fingerprint failures during pregnancy-related clinic visits9. As Bansal & Agrawal (2025) note, India’s digital ID infrastructure has become “more a barrier than an enabler” of social benefits for marginalized groups. In sum, algorithmic governance in welfare risks perpetuating existing inequalities: the poorest and least digitally literate can be quietly locked out of entitlements by a faceless system, with little remedy.
INFORMATIONAL PRIVACY UNDER INDIAN LAW
India’s Constitution now recognizes privacy as fundamental (under Article 21), but the law has struggled to keep pace with technology. The Puttaswamy judgment mandates that any state intrusion on personal data must satisfy strict tests of legality, necessity, and proportionality. Justice Chandrachud explicitly warned that advanced surveillance “technology and algorithmic governance” create new risks to dignity and autonomy10. In principle, citizens should enjoy a sphere of informational privacy where personal data cannot be arbitrarily collected, stored or shared.
Yet in practice the legal framework is weak. India lacks a comprehensive data-protection regime; the 2023 Digital Personal Data Protection Act is limited in scope and contains broad exemptions. Notably, the Act exempts key government agencies from its obligations in the interests of sovereignty, national security or public order11 (For example, intelligence and police agencies can escape independent oversight when invoking “security.”) Similarly, As per Malik, R. (2025), Section 69 of the Information Technology Act still grants sweeping powers of interception and monitoring without judicial warrant, a provision long criticized as overbroad. There are no statutory rules governing the use of algorithms by police or bureaucrats – no requirement to publish source code, conduct audits, or notify individuals affected. Thus, although privacy is a right on paper, state data practices have proliferated largely unchecked. Courts have begun to acknowledge these gaps: one petition against state face recognition argues there is no legal regulation for FRT (Malik, R., 2025). But so far judges have had few cases to establish oversight. In effect, the rise of AI tools is reconfiguring the risk calculus: pervasive data collection and opaque decision-making engines are now cutting into zones of personal information that once were presumptively protected.
EROSION AND THE NEED FOR NEW SAFEGUARDS
The changes above reflect a “quiet erosion” of informational privacy – gradual and technical, rather than through a single dramatic law change. Because these systems operate under the radar, public awareness and debate lag behind reality. Many citizens simply do not realize how much data the state holds or how algorithms influence their lives. At the same time, marginalized communities bear the brunt: poor and vulnerable individuals often lack the resources to challenge an errant algorithm, even as they suffer exclusions or profiling. This creates a chilling effect on privacy-sensitive activities (speech, association, dissent) even without overt censorship.
To arrest this erosion, scholars and activists urge a reconfiguration of legal safeguards that accounts for algorithmic governance. Key proposals include:
• Algorithmic Impact Assessments: Mandate ex-ante assessments for any high-risk public-sector algorithm (in welfare, policing, surveillance, etc.). Statutory AIAs would force agencies to evaluate the legality, fairness and privacy impact of a system before deployment. Such a requirement embeds “legality and proportionality” into the process and can reduce arbitrary or biased outcomes12.
• Transparency and Audits: Maintain public registries of government AI systems and require binding audits of data and models. For example, independent audits or “white-box” reviews could check that facial-recognition databases meet accuracy standards. Citizens should have a right to know when and how they are being profiled – perhaps via notices or technical explainers, and through strengthened Right-to-Information (RTI) rules.
• Oversight Bodies: Create an empowered Data Protection Authority or AI Commission with real independence. This body should have authority to scrutinize state algorithms, impose penalties for misuse, and order remedies. It could build expertise to advise courts or legislatures on technical matters (akin to judicial technology panels), ensuring that constitutional rights guide AI use.
• Individual Remedies: Ensure accessible redress channels. Affected individuals should be able to challenge automated decisions (e.g. welfare denials) with some form of evidentiary support and to demand human review where appropriate. Courts may need to develop doctrines (analogous to administrative law’s reason-giving requirement) that apply even to algorithmic decisions, so that agencies must document how an automated verdict was reached.
Implementing these principles would align India with emerging global norms (such as the EU’s AI Act). They echo calls by Indian experts for “algorithmic justice” – for instance, requiring mandatory AI impact studies and accountability documentation(Sonkar, 2025). Without such reforms, however, the net effect will be a significant shift in the balance of power: the state will increasingly see, record and act on citizens’ personal data with little constraint, effectively diluting the protective scope of Article 21.
CONCLUSION
State deployment of AI in India is accelerating across domains – policing, welfare, public health, and more – yet legal checks have not kept pace. This paper has shown how various AI-driven initiatives in India collect, integrate, and act on personal data in ways that chip away at informational privacy. What makes this particularly worrisome is its subtlety: without conscious legislative attention or high-profile litigation, technical systems are reconfiguring the boundaries of constitutional rights from within. Preserving privacy in this environment demands a scientific-tempered approach: one that grasps both the technical workings of algorithms and the social context of their use. Policymakers, jurists and the public must recognize that behind every data point is a person entitled to dignity. India’s constitutional safeguards need reinvigoration – via new laws, oversight mechanisms, and public discourse – to ensure that algorithmic governance serves citizens with accountability, not turns against them in silent ways.
*****
Footnotes
1. Christian Katzenbach and Lisa Ulbricht, ‘Algorithmic governance’ (2019) 8(4) Internet Policy Review. https://doi.org/10.14763/2019.4.1424
2. Malik, R. (2025). Cybersecurity and national security: Constitutional issues in digital governance. International Journal of Advance Research, Ideas and Innovations in Technology, 11(2), 165-169. https://www.ijariit.com/manuscripts/v11i2/V11I2-1265.pdf
3. Mahapatra, S. (2021). Digital surveillance and the threat to civil liberties in India. GIGA Focus Asia, (3). https://www.giga-hamburg.de/en/publications/giga-focus/digital-surveillance-and-the-threat-to-civil-liberties-in-india
4. George, T., & Sinha, A. (2025, July 1). As AI took over policing in Delhi, who bore the brunt? Pulitzer Center. https://pulitzercenter.org/stories/ai-took-over-policing-delhi-who-bore-brunt
5. George, T., & Sinha, A. (2025, July 1). As AI took over policing in Delhi, who bore the brunt? Pulitzer Center. https://pulitzercenter.org/stories/ai-took-over-policing-delhi-who-bore-brunt
6. Amnesty International. (2024a, April 12). Entity resolution in India’s welfare digitalization. Amnesty International. https://www.amnesty.org/en/latest/research/2024/04/entity-resolution-in-indias-welfare-digitalization/
7. Tapasya, K. S., Sambhav, K., & Joshi, D. (2024, January 24). How an algorithm denied food to thousands of poor in India’s Telangana. Al Jazeera. https://www.aljazeera.com/economy/2024/1/24/how-an-algorithm-denied-food-to-thousands-of-poor-in-indias-telangana
8. Mahapatra, S. (2021). Digital surveillance and the threat to civil liberties in India (GIGA Focus Asia No. 3). German Institute for Global and Area Studies (GIGA). https://www.giga-hamburg.de/en/publications/giga-focus/digital-surveillance-and-the-threat-to-civil-liberties-in-india
9. Bansal, P., & Agrawal, R. (2025, July 23). India’s digital divide and the consequent welfare bias. Eco‑Business. https://www.eco-business.com/opinion/indias-digital-divide-and-the-consequent-welfare-bias/
10. Malik, R. (2025). Cybersecurity and national security: Constitutional issues in digital governance. International Journal of Advance Research, Ideas and Innovations in Technology, 11(2), 165-169. https://www.ijariit.com/manuscript/cybersecurity-and-national-security-constitutional-issues-in-digital-governance/
11. DLA Piper. (2024). Data protection laws in India – Data Protection Laws of the World. DLA Piper. https://www.dlapiperdataprotection.com/?t=law&c=IN
12. AI Now Institute, Algorithmic Impact Assessments: Toward Accountable Automation in Public Agencies (2018).