Loading…
The IndiaAI Mission, approved by the Union Cabinet in March 2024 with a five-year outlay of ₹10,371 crore, comprises seven pillars — IndiaAI Compute, Foundation Models, AIKosh, IndiaAI Application Development Initiative, FutureSkills, Startup Financing and Safe & Trusted AI — Electronics & IT Minister Ashwini Vaishnaw informed Rajya Sabha on 24 July 2026.
Under the Safe & Trusted AI pillar, 13 Responsible AI projects have been approved across institutions like IIT Jodhpur, IIT Madras, IIT Delhi, IIT Kharagpur, NIT Raipur and DIAT, covering bias mitigation, machine unlearning, privacy-preserving AI, explainability and deepfake detection.
The AI Safety Institute (AISI), announced in January 2025, oversees these projects through a hub-and-spoke model and anchors India's risk-based AI Governance Guidelines, which propose institutional mechanisms — AIGEG, TPEC and AISI itself.
Mission-wide progress so far: 58 AI Centres of Excellence, 27 India Data & AI Labs (2,500+ students trained, 188 more underway), 686 fellowships across 178 institutions, over 26 lakh YUVA AI for All completions, and 93 lakh GPU hours sanctioned across 237 compute projects.
As Chair of the India AI Impact Summit 2026 in New Delhi, India produced the New Delhi Frontier AI Impact Commitments, a Guidance Note on AI Governance, and the Trusted AI Commons repository, while continuing to engage GPAI, G20 and the UN on global AI governance.
Build a comprehensive, inclusive AI ecosystem in India across compute, models, data, applications, skilling, startup funding and safety
Key: Seven pillars, ₹10,371 crore outlay over FY2024-29, implemented by MeitY
Promote responsible development, deployment and adoption of AI through indigenous governance frameworks, standards, tools and evaluation mechanisms
Key: 13 approved Responsible AI projects; funds work such as Saakshya (IIT Jodhpur & IIT Madras, deepfake detection), AI Vishleshak (audio-visual forgery detection) and IIT Kharagpur's real-time voice deepfake detector
National centralised repository of AI datasets, pretrained models and toolkits to democratise access to AI-ready data
Key: Launched 6 March 2025; now hosts over 15,000 datasets, 300+ AI models and 30+ toolkits across health, agriculture and education domains
Lay down a risk-based governance framework addressing algorithmic bias, misinformation, deepfakes and unintended societal harm
Key: Proposes three institutional mechanisms: AI Governance and Economic Group (AIGEG), Technology and Policy Expert Committee (TPEC), and the AI Safety Institute (AISI)
The planned successor to the Information Technology Act, 2000, expected to introduce risk-based platform classification and specific provisions on deepfakes and algorithmic transparency
Key: Remains in draft/public-consultation stage as of 2026 — not yet enacted; distinct from and not one of the IndiaAI Mission's seven pillars
India's dedicated body for AI safety, security and stakeholder trust; engages academia, startups, industry and Government via a hub-and-spoke model; oversees Safe & Trusted AI pillar projects
Multi-stakeholder international initiative for responsible, human-centric AI development; India is a founding member (joined June 2020) among 15 original members; now integrated with the OECD, taking membership to 44 countries
India's current primary digital-law statute; the Digital India Act, still in draft stage as of 2026, is expected to be the eventual statutory anchor for AI-specific obligations like deepfake labelling and algorithmic transparency that the Safe & Trusted AI pillar's guidelines currently address non-statutorily
Machine unlearning (the IIT Jodhpur project under Safe & Trusted AI) is the technique of removing specific unwanted, sensitive or outdated information from an already-trained AI model without retraining it from scratch — important when a model has memorised private or harmful data. Deepfake detection tools like Saakshya use a 'multi-agent' approach: several specialised AI models each check a different signal (facial artefacts, audio-visual sync, voice patterns) and combine their verdicts, which is more robust than a single detector.
Simple Analogy: Machine unlearning is like redacting a page from a book already printed, rather than reprinting the whole book. Multi-agent deepfake detection is like a panel of specialist doctors each examining a different symptom before giving a joint diagnosis.
GS Paper 3 > Science & Technology > AI, Robotics; GS Paper 2 > Governance > E-governance and Institutional Mechanisms
Consider the following statements: 1. Other than those made by humans, nanoparticles do not exist in nature. 2. Nanoparticles of some metallic oxides are used in the manufacture of some cosmetics. 3. Nanoparticles of some commercial products which enter the environment are unsafe for humans. Which of the statements given above is/are correct?
Answer: 2 and 3
High and rising — AI governance is one of the most actively tested current-affairs themes in 2025-26
IndiaAI Mission's national repository of AI datasets, pretrained models and toolkits, launched 6 March 2025
India's hub-and-spoke body for AI safety and security research, announced January 2025, overseeing Safe & Trusted AI pillar projects