The IndiaAI Mission represents the transition of e-governance from passive digitisation to active intelligence. But without addressing some crucial issues, AI in governance risks deepening existing inequities.” Comment.

When the Cabinet approved the IndiaAI Mission in March 2024 with a ₹10,371 crore outlay, it marked a genuine inflection — from e-governance as a digitisation exercise to e-governance as a decision-making architecture. The question is whether the infrastructure being built serves every citizen or accelerates the sorting of citizens into the governed and the ungoverned.

Passive digitisation to active intelligence

• IndiaAI Mission deploys AI for:
  • Predictive welfare delivery,
  • crop insurance claim processing under PM Fasal Bima,
  • health diagnostics under ABDM,
  • judicial pendency reduction under e-Courts Phase III.
• Fraud detection and DBT leakage prevention:
  • AI algorithms now actively flag duplicate beneficiaries, ghost entries, and pattern anomalies in real time rather than waiting for audit reports. Ex: PM-KISAN
• Predictive policing and crime pattern analysis:
  • Police forces are moving from reactive FIR registration to predictive deployment based on crime pattern modelling.
• Natural language grievance routing:
  • CPGRAMS upgraded with AI-based categorisation now reads grievance text, identifies the responsible ministry, routes automatically, and flags overdue cases — eliminating the passive queue model where grievances sat undifferentiated.
• Tax compliance from retrospective audit to real-time intelligence:
  • Faceless Assessment under Income Tax uses AI to match TDS data, GST returns, financial statements, and third-party data in real time, flagging discrepancies before they become evasion.

Crucial issues which need resolution:

• Data privacy without structural protection:
  • DPDP Act 2023 grants government agencies broad exemptions on "national security" grounds without a proportionality requirement.
  • Ex: NATGRID aggregates data from banking, immigration, telecom, and tax systems with no legislative backing and no oversight mechanism, it fails Puttaswamy's (2017) legality test at threshold.
• Algorithmic accountability is absent by design: IndiaAI Mission's Safe & Trusted AI pillar selected 8 projects from 2,000+ proposals covering bias mitigation and explainability, but these remain R&D investments, not regulatory mandates.
  • Ex: No law currently requires AI systems used in government welfare delivery to be auditable or appealable.
• Digital divide compounds the exclusion: Only 2.7% of households in the poorest quintile have computer access (Oxfam 2022).
  • Ex: BharatNet connected 2,14,283 of 2,22,343 targeted Gram Panchayats by October 2024 — last-mile connectivity remains elusive precisely where AI-powered welfare delivery is most needed.

Active intelligence that cannot reach the last mile is not a governance upgrade, it is a governance subsidy for those already connected. The IndiaAI Mission will fulfil its democratic promise only when its Safe & Trusted AI pillar becomes a legally enforceable framework, and its compute capacity reaches the citizen who has neither a smartphone nor a broadband connection.