Responsible AI for Bharat: Trust, Safety, and Global Leadership

Contents

Executive Summary

This inaugural panel at the AI Impact Summit discusses India's strategic transition from AI consumer to creator, emphasizing inclusive, frugal innovation and trustworthy AI as a differentiator for the Global South. The discussion positions India as a potential global leader in responsible AI by embedding governance from inception, developing context-specific models, and democratizing AI access across 1.4 billion people—contrasting with Western approaches focused on frontier model scale and capability.

Key Takeaways

India's AI Strategy Must Be Locally Rooted: Success means building contextual, multilingual models trained on Indian data for Indian problems (health, agriculture, governance, education)—not scaling Western foundation models.
Trustworthy AI Is a Competitive Advantage, Not a Cost: Embedding fairness, transparency, and governance from day one positions India as the global leader in responsible AI and blueprints the approach for developing nations.
Education Transformation Is as Critical as Model Building: Producing 600 million AI creators (not just users) requires reimagining education, leveraging AI for personalized learning, and building AI literacy into curricula from schools to universities.
Standards-Setting Authority Requires Participation: India must contribute to—not just adopt—global AI governance frameworks to prevent Western standards from becoming developmental barriers. The Delhi Declaration should encode India's commitment to contextual, inclusive standards.
Frugal, Inclusive AI Innovation Is Scalable Innovation: Optimizing for efficiency, local languages, diverse literacy levels, and lower compute costs is not a niche approach—it's the model for reaching 1.4 billion people and the Global South at population scale.

Key Topics Covered

India's Economic and Strategic Positioning: India's potential GDP trajectory (70-170 trillion by some projections) and role as a leader for the Global South
From Consumer to Creator: Shifting India's AI strategy from adopting Western foundation models to building indigenous, context-aware solutions
Frugal Innovation & Scalable Models: Leveraging DPI (Digital Public Infrastructure), efficient compute, and distributed approaches rather than competing on parameter scale
Trustworthy AI as Infrastructure: Embedding governance, fairness, and reliability at the design stage rather than as post-hoc compliance
Multilingual and Contextual AI: Ensuring fairness, accuracy, and performance across India's diverse languages, geographies, and literacy levels
Education and Skilling: Reimagining education systems to produce AI builders, not just users; leveraging India's 600 million young population and 40 million higher education institutions
Global Standards and Governance: India's role in shaping AI governance frameworks and preventing Western-defined standards from creating barriers for developing nations
Nation-Building with AI: Using AI as a tool for inclusive development, healthcare, agriculture, and government efficiency in the Global South context
Capital and Investment: Addressing the severe underfunding of trustworthy AI and responsible AI solutions versus frontier model development
Geopolitical Considerations: Balancing collaboration with frontier AI companies while protecting India's developmental autonomy and standards-setting authority

Key Points & Insights

Context Matters Over Scale: Western models trained on different data, contexts, and languages won't work effectively for India's complex, multilingual, geographically diverse population. India needs indigenous models optimized for local conditions, not larger parameter counts.
DPI as an Enabler: Digital Public Infrastructure (Aadhaar, UPI) demonstrated India's capacity to build scalable, inclusive backbone systems. Similar infrastructure frameworks are needed for AI—data ecosystems, computation layers, and open standards.
Governance as First Principle, Not Afterthought: Trust and safety must be embedded during AI system design, not bolted on later. This includes understanding AI supply chains, evaluating third-party dependencies, testing for fairness across diverse populations, and measuring trust scientifically.
AI Humanization Requires Local Knowledge: LLMs interface with users in everyday language, making them inherently humanized. But what feels "human" and trustworthy varies dramatically across cultural contexts—Maharashtra village users vs. Bay Area users. Continuous benchmarking against local experience is essential.
Frugal Innovation ≠ Compromise on Quality: Efficient resource use, lower compute costs, and smaller models are not lower-tier alternatives—they are strategic advantages for scaling AI to 1.4 billion people. India can and should compete globally while optimizing for resources.
Education System Redesign is Critical: Traditional 200-year-old education systems built for the industrial age must shift toward personalized, adaptive AI-driven learning where students learn in ways that suit their skills. 600 million young Indians need reskilling, not just curriculum insertions.
Conflicting Standards as a Barrier: As the Global South develops AI, Western-defined governance and compliance standards may emerge as regulatory obstacles. India must actively participate in creating standards, not just adopting them, to avoid developmental hindrance.
Capital Misdirection: Less than 0.001% of AI-focused capital globally goes to trustworthy AI and measurable trust solutions; billions go to frontier models. VCs and startups should invest heavily in responsible AI infrastructure, benchmarks, and guardrails.
Human Flourishing as the North Star: Rather than "human-in-the-loop," the framework should be "human-in-the-center"—AI should serve human welfare, equitable access, and common good, not maximize corporate or state power.
Geopolitical AI Will Persist, But Process Leadership Matters: The AI competition for global supremacy won't disappear, but India can lead the process of inclusive AI development and ensure the Global South participates in governance-setting, not just consumption.

Notable Quotes or Statements

On India's Economic and Strategic Role

"India's potential by open standard around 70 trillion plus the potential by American standard almost 170 trillion even more than the current size of global economy 120 trillion so what I'm talking about is not for today's India but for India of tomorrow the future of India the future of AI will be drafted and created by these numbers." — First speaker (Suresh Yadav, Intel)

On Trustworthy AI

"If you're going to think about governance and trustworthiness as the last step in building AI, I think India will not have the opportunity to win in this AI arms race... I think the way India is going to lead is by focusing on trustworthy AI principles... making it work for the massive population that exists in India." — Navina Singh

On Multilingual Fairness

"The interface between us and these models is everyday language. So by definition they are very humanized in some sense... What is humanizing varies from people to people. What somebody in a village in Maharashtra feels versus what somebody feels in western Europe or in Bay Area would be very different." — Pratush Kumar (implicit speaker on LLM/fairness)

On Education as AI Infrastructure

"Education is 200 year old system that has been built. It is built for the industrial age and I think it's slowly adapting to the information age... How can AI go back and change fundamental ways in which we do things in this society..." — Santosh Viswanathan

On Human-Centered Design

"The fundamental shift which I see... people talk about human in the loop but here it give the summit give the concept of human in the center so that's the fundamental shift which I see and human in the center is going to make responsible AI by action by impact." — Suresh Yadav

On Standards and Sovereignty

"When the west develops a model, there are no governance standards. There are no risk frameworks. They develop and it goes to the market to the global market. But when the global south will start developing lot of new standards, lot of new barriers will come. So that is where the governments have to be cautious that which standards developed by whom are you part of the process of developing that standards." — Suresh Yadav

On Global South Leadership

"India being that promised land where the new trajectory of artificial intelligence will kick off... India under Prime Minister Modi is that voice of that global south which will ensure that that there is no have and have nots." — Minister of State Mithun Prasada (implicitly quoted)

Speakers & Organizations Mentioned

Government & Policy

Mithun Prasada — Minister of State for Commerce & Industry, and Electronics & IT (India)
Prime Minister Narendra Modi (referenced)
Professor Sud — Science Adviser to Prime Minister Modi

Panelists (Identified or Strongly Implied)

Suresh Yadav — Intel (opening speaker on DPI and frugal innovation)
Navina Singh — Governance and trustworthy AI expert
Pratush Kumar — Foundation model builder (Google, multilingual LLM work implied)
Santosh Viswanathan — Education and skills transformation (implied speaker on UPI/education analogy)

Companies & Institutions Referenced

Intel (AI Chopal initiative mentioned)
Google (frontier model provider)
Anthropic (frontier model provider, mentioned as present at summit)
OpenAI (frontier model provider, mentioned as present at summit)
Indian government institutions (various)

Event/Venue

AI Impact Summit (New Delhi, India) — inaugural day
Delhi Declaration (expected outcomes by summit end)

Technical Concepts & Resources

AI Infrastructure & Architecture

DPI (Digital Public Infrastructure): Aadhaar, UPI cited as successful backbone models; framework for scaling AI systems
Foundation Models: Western frontier models (GPT-class scale) vs. India-specific smaller, contextual models
Agentic Approaches: Mentioned as enhancement to existing datasets for model building
AI Supply Chain Transparency: Software bill of materials analogy; vendor dependencies and third-party model providers must be mapped and audited
Compute Optimization: Focus on frugal compute, efficient data centers, lower GPU costs as strategic priorities

Fairness, Safety & Governance

Multilingual Fairness Benchmarks: Regional/linguistic accuracy and user experience evaluation beyond English-centric metrics
Trust Measurement: Scientific frameworks for measuring trust (not just compliance checklists)
Contextual Evaluation Methodologies: Benchmarks and datasets tailored to Indian demographics, languages, and use cases (not transplanted Western fairness benchmarks)
Impact Assessment: Measuring consequences of AI systems on 1.4 billion diverse users across sectors (health, agriculture, government)
Third-party Vendor Risk: Understanding dependencies on multiple foundation model providers and their governance/training practices

Data & Datasets

Indian Language Datasets: Public datasets made available by India for experimentation, testing, and training
Sector-Specific Data: Agriculture, health, governance data for training domain-specific models
Evaluation Datasets: Needed for linguistic fairness, cultural context validation, reliability testing

Education & Skilling

Personalized AI-Driven Learning: Adaptive curriculum based on individual student needs (vs. fixed, one-way curriculum)
AI Literacy Programs: Including governance, trustworthiness, fairness, compliance, reliability concepts—not just model-building
Bottom-Up Language-Breaking Approaches: Using web-driven systems to reach non-English speakers and non-literate populations

Policy & Governance Frameworks

India's AI Governance Framework: Technology-focused, context-grounded approach (referenced as distinct from other regulatory models)
Standards Development Process: India's participation in creating contextual standards vs. adopting Western standards
Scalable Solutions for Sectors: Agriculture, healthcare, government, education—sector-specific AI applications

Strategic Concepts

Frugal Innovation: Resource optimization without compromising quality or scale
Inclusive AI: Population-scale accessibility, multilingual support, diverse literacy and educational backgrounds
Nation-Building with AI: Using AI for developmental goals (reducing poverty, improving governance, healthcare access, agricultural productivity)
Creator vs. Consumer Mindset: Shift from adopting global models to building, innovating, and contributing to global AI landscape

Context & Relevance

This panel represents a pivotal moment in India's AI strategy positioning: a deliberate departure from the assumption that Indian AI success means adopting or scaling Western frontier models. Instead, speakers articulate a vision where India's complexity, diversity, and developmental challenges become assets—forcing innovation in fairness, multilingualism, efficiency, and governance that naturally serves the Global South and, by extension, contributes blueprints for equitable AI globally.

The emphasis on "human-in-the-center" and trustworthy AI-as-first-principle reflects a philosophical repositioning of AI from a competitive arms race (in which India would lose on parameter scale) to a leadership position in responsible scale—proving that inclusive, efficient, and trustworthy AI is not a compromise but a differentiator.