Trusted AI for Everyone: USISPF Panel on Global AI Impact

Contents

Executive Summary

This panel discussion at India's AI Summit focused on scaling AI access to 8 billion people globally while ensuring trust and preventing a widening digital divide between the Global North and South. The panelists emphasized that while AI infrastructure investment is critical, India's competitive advantage lies in AI applications, developer talent, and localized solutions rather than competing on LLMs or hardware.

Key Takeaways

AI Divides Are Reversible, But Only Through Intentional Ecosystem Building: The Global North-South adoption gap (25% vs. 14%) is accelerating, not improving. Closing it requires simultaneous investment in infrastructure, skilling, localized applications, and demand stimulation—not one or the other.
India's Realistic Path to AI Leadership Is Applications, Not Foundations: India cannot outspend the US and China on LLMs and GPUs. But with 1.4 billion people, mobile-first digital behavior, and deep domain expertise across agriculture/healthcare/finance, India can dominate the application layer—where actual economic value and jobs are created.
Language and Interface Design Are Democracy or Exclusion: Ensuring AI works in Hindi, Tamil, Marathi, etc., and operates via voice (not typing) is not a "nice-to-have" feature—it's the infrastructure of equitable AI. Current Western-centric models automatically exclude billions.
Trust Is an Engineering Problem, Not Just a Governance Problem: Companies like Zoom can refuse to train on customer data without legal mandate; Rubric can build guardrails to prevent hallucinations in critical sectors. Responsible AI practices often require technical discipline before regulation.
This Moment Favors Risk-Takers and Job Creators: Cultural shift in India (80% of IIT/IIM grads now start companies; failure is acceptable) combined with abundant US capital and government infrastructure (AI Kosh, 7,500+ datasets, subsidized compute) creates a rare window for the next 10,000 startups to emerge in AI applications.

Key Topics Covered

The AI Divide: Disparity in AI adoption between Global North (25% of working-age population) and Global South (14%)
Infrastructure as Foundation: Data centers, connectivity, and electricity required for equitable AI access
Skilling and Education: Critical need to train developers and educators across emerging markets
Multilingual AI: Ensuring AI works equally well in all languages, not just English
India's Strategic Position: India's role as AI adoption leader, application developer, and global talent exporter
Trust and Safety in AI: Building guardrails, deterministic responses, and responsible AI practices
Economic Opportunity: AI as path to catch-up growth for Global South; abundance economy vs. job displacement
Data Ownership and Equity: Questions about who benefits from data used to train models
Regulatory Approach: "Innovation-light" regulation in India vs. restrictive European approaches
Localized Solutions: Importance of voice-first, multilingual applications for rural and underserved populations
Startup Ecosystem and Funding: Role of US capital in funding Indian startups; cultural shift toward risk-taking
Government Infrastructure: India's AI Kosh program providing subsidized compute and open datasets

Key Points & Insights

The 25-14 Gap Is Urgent: Microsoft's AI diffusion report shows 25% adoption in the Global North vs. 14% in the Global South. Growth rates are diverging further (1.8% vs. 1.0%), making this the defining inequality of the AI era.
Infrastructure = Prerequisite, Not Guarantee: Brad Smith outlined the formula: data centers + connectivity + electricity must come together. Microsoft's $170 billion India investment and $50 billion Global South commitment are necessary but insufficient without demand generation.
India's Non-LLM Competitive Advantage: All panelists agreed India cannot compete on large language models or GPUs but can dominate in AI applications, adoption, and deployment. This is where job creation and economic value actually materialize.
Multilingual AI Is Non-Negotiable for Equity: Current AI models favor English speakers. Bhashini and other multilingual initiatives are essential infrastructure; models trained on narrow language datasets perpetuate exclusion of the 1.4+ billion Indians who don't use English daily.
Voice-First, Not Keyboard-First, Design: S. Krishnan emphasized that India's digital access pattern (mobile-phone-first) means AI must be voice-driven and work across 22+ Indian languages and dialects. Western UI/UX assumptions don't transfer.
Trust Requires Deterministic Guardrails, Not Just Promises: Bipul highlighted that AI hallucinations destroy trust in critical sectors (healthcare, finance). Fine-tuning models, creating guardrails, and making AI outputs verifiable are engineering challenges, not regulatory workarounds.
Data as Leverage, Not Just Input: Several speakers pointed out that without Global South data, big tech's $150+ billion capex cannot generate ROI. India has negotiating power—Aperna noted Zoom's decision not to train on customer data without regulatory mandate, showing voluntary responsible practices are possible.
The Talent Export Will Dominate the Next Decade: Umesh noted Indian-Americans/Asians are concentrated in AI company leadership. India is and will be the largest talent exporter for AI; if that talent learns AI skills at scale, India's economic growth could accelerate sharply.
"Age of Abundance" vs. Job Displacement Narrative: Umesh and Bipul reframed job displacement fears: if India adopts AI at scale, GDP growth will sharply rise, creating more jobs in applications, deployment, and maintenance. Y2K moment analogy suggests abundance, not scarcity.
Small Models for Real Problems: S. Krishnan clarified that India doesn't need only large foundational models. Agriculture, healthcare, manufacturing problems can be solved with smaller, quantitative, or vision models—lowering the compute and data barriers to innovation.

Notable Quotes or Statements

Brad Smith (Microsoft): "AI is either going to close the great divides that exist in the world or it will make those divides even wider. And I think that is the reality we must confront head on."
Brad Smith: "The greatest divide that exists on our planet is the economic divide between the global north and the global south, which was fundamentally created by unequal access to technology, specifically electricity."
Aperna Bawa (Zoom): "There can be no hidden logic...you as a user should be able to use our product and not have to figure something out with a PhD degree."
Bipul (Rubric): "Where do we fit in as human beings? We fit in to do intuition work where we connect new dots that has not been connected yet."
Umesh Sajdev (Unifor): "AI is going to lead us to an age of abundance. India's path to cross 10% growth GDP will be because of AI...that path of abundance will create more jobs, not less."
Dr. Agi (USISPF): "The question is not whether AI will scale, because it will. The question is can we scale trusted AI?"
Dr. Agi: "What we are seeing is a feudalism in AI—few companies will control both hardware and LLMs. India's opportunity is the app side."
Umesh Sajdev: "Let there be the next 10,000 Unifors in the next 5 years globally. For every sovereign local LLM, there should be 5,000 AI application companies."
S. Krishnan (MeitY Secretary): "Our embrace of AI is innovation first...the opportunity to become a Viksit Bharat, a developed nation by 2047, is by riding this wave of technology."
S. Krishnan: "It's probably the Y2K moment all over again. India is going to lose far fewer jobs than the developed world."
Aperna Bawa: "If you can make things super simple, super easy...that increases GDP for the country."

Speakers & Organizations Mentioned

Speaker	Title/Role	Organization
Dr. MKkesh Agi	President & CEO	US-India Strategic Partnership Forum (USISPF)
Brad Smith	Vice Chairman & President	Microsoft
Aperna Bawa	Chief Operating Officer	Zoom
Umesh Sajdev	CEO & Co-Founder	Unifor
Bipul	CEO, Chairman & Co-Founder	Rubric
S. Krishnan	Secretary	Ministry of Electronics & Information Technology (MeitY), India
Satya Nadella	(mentioned, not present)	CEO, Microsoft
Jensen Huang	(mentioned, not present)	CEO, Nvidia

Other Entities Mentioned:

US-India Strategic Partnership Forum (USISPF)
Microsoft
Zoom
Unifor
Rubric
OpenAI
IIT Madras
Bharat Mandapam (summit venue)
Government of India / Ministry of Electronics & Information Technology (MeitY)
Perplexity (AI company)
Lens Cart, Bajage (Indian companies)

Technical Concepts & Resources

AI Models & Platforms

ChatGPT: Referenced as democratizing AI through natural language; released 3 years and 3 months before this talk
Generative AI: Focus on diffusion and equitable access
Sovereign/Local LLMs: Models developed by nations/regions rather than US/China dominance
Small Models: Emphasized for domain-specific problems (agriculture, healthcare) vs. foundation models
Vision Models & Quantitative Models: Alternatives to language models for specific applications

Infrastructure & Datasets

Bhashini: Government multilingual AI initiative supporting 22+ Indian languages and dialects; voice-first interface
AI Kosh: India's subsidized compute infrastructure for startups; provides access to compute at 1/3 global cost
Datasets: India's AI Kosh provides 7,500+ open datasets (grown from 300 six months prior); ongoing expansion
Data Centers: Critical bottleneck; Microsoft investing $170 billion in India, $50 billion in Global South

Methodologies & Practices

Fine-tuning Models: Approach used by Rubric to ensure deterministic, trustworthy outputs
Guardrails: Technical controls preventing hallucinations and off-policy responses
Agentic AI: Autonomous AI systems; mentioned in context of Unifor's platform
Voice-First Design: Replacing keyboard-dependent interfaces for accessibility in India
Multilingual NLP: Critical for serving non-English speakers; current gap in Western models

Key Metrics & Data Points

AI Adoption Gap: 25% Global North vs. 14% Global South (Microsoft AI Diffusion Report, end 2025)
Growth Rate Divergence: 1.8% in Global North vs. 1.0% in Global South (H2 2025)
Microsoft Capex: $170 billion in India; $50 billion target in Global South by end of decade
Data Availability: 7,500+ datasets in AI Kosh; target to grow much higher
Compute Cost: 1/3 of global average through AI Kosh
Talent Pipeline: ~20% of world's digital workers will be from India by 2030
India's Population: 1.4 billion; 80% of recent IIT/IIM graduates now start companies

Government & Policy Frameworks

"Innovation-Light" Regulation: India's approach contrasted with Europe's regulatory-first (AI Act)
AI Kosh: Government program democratizing access to compute, models, and datasets
Bhashini Initiative: Multilingual language support across government services
Y2K Analogy: Government official's comparison for scaling talent and opportunity

Context & Significance

This panel represents a strategic moment in global AI governance: major US tech companies (Microsoft, Zoom) are publicly committing to equitable Global South access, while India's government is positioning itself as the "AI application and adoption" leader rather than attempting to compete with US/China on foundational models. The emphasis on voice-first, multilingual, small-model-based solutions is pragmatic acknowledgment that "Western" AI won't serve 1.4 billion Indians without redesign. The recurring theme of opportunity over threat (jobs will be created, not destroyed; India can catch up if it acts now) is strategically important for public buy-in, though not universally agreed upon in AI safety discourse.