How AI Is Being Used in India's Government
India's government is deploying artificial intelligence across an expanding range of public sector functions — from crop advisory systems that predict pest outbreaks to judicial AI tools that summarise case files, from CERT-In's AI-driven threat detection to Doordarshan Kisan's multilingual AI news anchors.
The IndiaAI Mission's "AI for India" pillar specifically targets four application domains: healthcare (AI diagnostics for cancer, tuberculosis, retinal disease), agriculture (crop advisory, yield prediction, pest and disease early warning), education (personalised learning systems, automated grading), and governance (public service delivery optimisation, fraud detection in welfare schemes).
The government has established AI Centres of Excellence (CoEs) in Healthcare, Agriculture, and Sustainable Cities in New Delhi; Budget 2025–26 announced a fourth CoE for AI in Education with ₹500 crore outlay.
 |
| Representational Image: How AI Is Being Used in India's Government |
This AI targeting intersects directly with India's data protection framework — the DPDPA's automated decision-making provisions (which require explainability for significant decisions) apply to AI systems that determine welfare eligibility, creating a potential conflict between efficiency-focused AI and rights-protective data law.
The Supreme Court's Puttaswamy judgment's privacy
foundation implies that AI systems making consequential decisions about
citizens' welfare access must satisfy proportionality and accountability
requirements — but the specific regulatory framework for government AI remains
under development.
What You Need to Know
- IndiaAI
Mission CoEs: AI Centre of Excellence for Healthcare (disease detection
including cancer, TB, diabetic retinopathy); AI CoE for Agriculture (crop
advisory, AgriStack integration, pest early warning); AI CoE for
Sustainable Cities; new AI CoE for Education (Budget 2025, ₹500 crore);
these CoEs are public-private partnerships developing India-specific AI
applications.
- Judicial
AI: Supreme Court's SUVAS (Supreme Court Vidhik Anuvaad Software)
translates court judgments into regional languages; SUPACE (Supreme Court
Portal for Assistance in Courts Efficiency) provides AI research
assistance for judges; National Judicial Data Grid uses AI analytics for
case flow management; at district court level, AI pilot programmes for
e-FIR processing and bail assessment are underway in some states.
- CERT-In
AI for cybersecurity: WEF Global Cybersecurity Outlook 2025 highlighted
CERT-In's AI-driven situational awareness systems for detecting malicious
domains and phishing; I4C uses AI for fraud pattern detection in
cybercrime; AI threat intelligence is integrated into CERT-In's national
monitoring infrastructure.
- Doordarshan
Kisan AI anchors: DD Kisan launched AI anchors "Krish" and
"Bhoomi" delivering agricultural news in 50 languages; India's
first government TV channel to use AI presenters; represents an early
deployment of AI in government broadcasting.
- AgriStack
and AI agriculture: Digital Agriculture Mission (₹2,817 crore) integrates
farmer digital IDs (Aadhaar-linked for 11 crore farmers), Bhu-Aadhaar land
records, and crop insurance data into a unified platform enabling
AI-driven personalised advisory; 19 states have signed MoUs for AgriStack
as of 2024.
How It Works in Practice
1. AI in welfare fraud detection: The government's
DBT monitoring systems use AI to flag anomalies in beneficiary data — duplicate
Aadhaar numbers, suspicious clustering of beneficiaries at single addresses,
bank accounts receiving payments without biometric authentication. These AI flags
trigger human review and potential suspension of benefits. Civil society
organisations have documented cases where legitimate beneficiaries were flagged
and excluded by AI systems based on patterns that turned out to be legitimate;
the "explainability" of these AI decisions — why a specific
beneficiary was flagged — is not systematically available to affected
individuals.
2. AI crop advisory and the Digital Agriculture Mission:
ICAR (Indian Council of Agricultural Research) and state agriculture
departments are deploying AI-powered crop advisory systems that combine
satellite imagery, soil data, weather forecasting, and historical crop
performance to provide personalised recommendations to farmers. The AgriStack
platform integrates these data sources; BHASHINI enables voice-based delivery
in local languages. AI crop advisories have the potential to significantly
improve agricultural productivity and reduce input costs; their effectiveness
depends on data quality and last-mile delivery to farmers who may not have
smartphones.
3. Judicial AI and access to justice: Supreme Court
AI tools (SUVAS, SUPACE) primarily serve efficiency goals — faster translation,
better research support — rather than replacing judicial decision-making. The
translation capability (SUVAS translating judgments into 18 regional languages)
directly addresses access to justice for litigants who cannot read English or
Hindi. Bail assessment AI pilots in some district courts are more controversial
— algorithmic inputs to bail decisions raise due process concerns that India's
judiciary has not fully examined.
4. Health AI diagnostics: India has approved AI
diagnostic tools for: diabetic retinopathy screening (AI can screen fundus
images with accuracy comparable to ophthalmologists); tuberculosis detection
(AI analysis of chest X-rays); and cancer screening support. These tools are deployed
in public health facilities as decision-support tools for healthcare workers;
they do not replace clinical judgment but can triage patients in settings where
specialist access is limited.
5. Smart cities and urban AI: India's 100 Smart
Cities Mission — which funds urban digital infrastructure — has produced AI
applications in traffic management, waste management monitoring, public safety
surveillance, and urban infrastructure monitoring. The constitutional
implications of urban AI surveillance — facial recognition at public spaces,
AI-enabled CCTV analysis — raise concerns about surveillance normalisation that
India's data protection framework does not yet specifically address.
What People Often Misunderstand
- Indian
government AI is predominantly decision-support, not autonomous
decision-making: Most deployed government AI tools assist human
decision-makers rather than making final decisions autonomously; the
"AI anchor" is genuinely autonomous delivery; welfare AI flags
are reviewed by humans; judicial AI provides research, not verdicts.
- AI
in agriculture faces last-mile delivery challenges: The sophistication
of AI crop advisory systems matters less than whether farmers actually
receive and act on the recommendations; delivery through BHASHINI voice
interfaces is more accessible than app-based interfaces but still requires
connectivity and digital literacy that many Indian farmers lack.
- Government
AI procurement lacks public transparency: How government AI systems
are procured, what their performance benchmarks are, how they are audited
for bias and accuracy, and what recourse citizens have when AI systems
produce wrong outcomes are questions that current Indian governance
frameworks do not systematically answer.
- Facial
recognition deployment in India is insufficiently regulated: Smart
city surveillance deployments using facial recognition — at airports,
transit hubs, and increasingly public spaces — operate without specific
legal authorisation, data retention limits, or independent oversight; the
DPDPA's biometric data protections nominally apply but the government
exemptions may significantly limit their effectiveness in this context.
- India's
AI applications are predominantly for lower-income government service
users: Unlike EU or US AI governance discussions centred on
consumer-facing private sector AI, India's government AI applications
primarily serve welfare beneficiaries, farmers, and the justice system —
populations with limited voice to challenge problematic AI outcomes.
What Changes Over Time
The AISI's development of safety testing standards — expected in late 2026 — will, if applied to government AI as well as private sector AI, create the first systematic evaluation requirements for public sector AI deployments.
The DPDPA's implementation (compliance deadline May
2027) will require government agencies using AI for automated decision-making
about citizens to assess explainability and bias implications.
Sources and Further Reading
- PSA
— AI Mission and Initiatives: https://www.psa.gov.in/ai-mission-initiatives
- Drishti
IAS — 10 Years Digital India: https://www.drishtiias.com/daily-updates/daily-news-analysis/10-years-of-digital-india
- PIB
— CERT-In 2025: https://www.pib.gov.in/PressReleasePage.aspx?PRID=2217537®=3&lang=1
- Truyo
— AI governance India: https://truyo.com/governing-the-ai-surge-how-india-is-writing-the-rulebook-for-responsible-ai/
- Protean
— DPI 2024: https://www-proteantech-in.translate.goog/articles/dpi-2-0-2-4-developments/