Published: Yojana, April 2026 (focus article by Abhishek Singh)

The issue’s central argument: India is repositioning artificial intelligence as the next layer of Digital Public Infrastructure (DPI), after Aadhaar, UPI and DigiLocker. The shift is framed as moving from “technology-centric AI to citizen-centric AI”, where the state builds shared AI rails that startups, researchers and governments ride, rather than leaving capability concentrated in a few private firms.

The IndiaAI Mission’s Three Layers

Layer What it provides
Compute Shared GPU capacity at subsidised rates of Rs 65-100 per GPU-hour
Data and models (AIKosh) National repository: 10,000+ datasets, 286 AI models
Applications Crop monitoring, disaster prediction, beneficiary verification, diagnostics (TB, diabetic retinopathy), voice-enabled UPI

The compute subsidy is the population-scale wager: by collapsing the cost of training and inference, India lowers the entry barrier the way UPI collapsed the cost of payments.

Governance Scaffolding

  • DPDP Act 2023: penalties up to Rs 250 crore; Data Protection Board as the adjudicatory body
  • IT Intermediary Guidelines (amended): deepfake content must be removed within 24 hours of reporting
  • The threat side: I4C data shows a 250 percent rise in deepfake-related crimes in 2024
  • Judiciary adoption: SUVAS (judgment translation across Indian languages) and SUPACE (research assistance for judges)

Diplomacy and Skilling

The India AI Impact Summit 2026 in New Delhi produced the New Delhi Declaration, endorsed by 90+ countries, positioning India as the voice of the Global South on AI governance: access to compute, datasets in local languages, and AI for development rather than only frontier-model safety.

Skilling rails: YUVA AI for ALL courses on DIKSHA (schools) and iGOT Karmayogi (civil servants), plus AI labs planned in Tier-2 and Tier-3 cities.

Mains Angle

For: DPI-style AI avoids private concentration, serves welfare delivery directly, and exports a replicable model to developing countries. Against / risks: subsidised compute can crowd out private data-centre investment; deepfake enforcement is reactive; DPDP’s state exemptions weaken the privacy counterweight that population-scale AI needs. Way forward: independent algorithmic audit capacity, compute-allocation transparency, and Indic-language dataset quality standards.

📌 Facts Corner — Knowledgepedia

IndiaAI Mission:

  • GPU compute subsidised at Rs 65-100 per GPU-hour
  • AIKosh: 10,000+ datasets, 286 AI models

Law and institutions:

  • DPDP Act 2023: penalties up to Rs 250 crore; Data Protection Board
  • IT rules: deepfakes must come down within 24 hours
  • I4C (Indian Cyber Crime Coordination Centre): 250 percent rise in deepfake crimes (2024)

Other relevant facts:

  • India AI Impact Summit 2026, New Delhi: New Delhi Declaration endorsed by 90+ countries
  • Judiciary AI tools: SUVAS (translation), SUPACE (court efficiency)
  • YUVA AI for ALL runs on DIKSHA and iGOT Karmayogi

Sources: Yojana / Publications Division, PIB