India AI Impact Summit Insights
All sessions

Building Trusted AI at Scale: Cities, Startups & Digital Sovereignty | India AI Impact Summit 2026

Contents

Executive Summary

The India AI Impact Summit 2026, held in New Delhi with approximately 250,000 attendees, positioned India as a central player in shaping global artificial intelligence policy, governance, and deployment. The summit moved beyond theoretical AI discussion to focus on practical adoption, responsible development, and ensuring AI serves human dignity—with particular emphasis on how the global south, emerging economies, and India specifically can lead inclusive AI development while maintaining sovereignty and ethical standards.

Key Takeaways

  1. AI is entering phase 2 (agents) and moving toward phase 3 (physical AI)—the pace of change is accelerating beyond human absorption capacity, requiring fundamental shifts in how organizations deploy and govern technology.

  2. Infrastructure, context, and trust are the three critical constraints that must be solved simultaneously; solving any one without the others is insufficient for safe, beneficial AI deployment at scale.

  3. India must build for India and the Global South—leveraging its scale, talent, and development challenges to create sovereign, domain-specific AI systems that solve local problems while solving global ones, rather than becoming a consumer of Western AI.

  4. Human responsibility and judgment cannot be automated away, especially in defense, healthcare, and finance—guardrails, oversight, and accountability must be built into every AI system from the start, not retrofitted as an afterthought.

  5. Skills and jobs will transform, not disappear—workers must shift from individual task-execution to orchestration and critical thinking; the industry, government, and philanthropy must invest in large-scale, continuous reskilling infrastructure.

Key Topics Covered

  • Three Core AI Constraints: Infrastructure scarcity, context gaps in AI systems, and trust deficits
  • Shift from Chatbots to Agents: Moving from question-answering systems to autonomous task-execution agents
  • Physical AI & Edge Computing: Deploying AI directly on devices rather than centralizing in cloud
  • Digital Sovereignty & Data Governance: National control over data, compute, and AI infrastructure
  • AI in Legacy Modernization: Using AI to address technical debt and modernize 15–20-year-old systems
  • Global AI Governance & Safety: Frameworks for responsible AI, human control in high-stakes decisions, and international standards
  • Skills & Employment: Job creation through AI rather than replacement; new role archetypes emerging
  • AI for Development: Healthcare, agriculture, education, financial inclusion in emerging economies
  • Defense & Autonomous Systems: Meaningful human control in military AI; ethical decision-making under pressure
  • Infrastructure & Compute: Power requirements, data center deployment, carbon efficiency, and network architectures (6G)
  • India's Demographic Dividend: India's young population and tech talent as unique advantage in AI era

Key Points & Insights

  1. Infrastructure as Oxygen for AI: The fundamental constraint to AI progress is insufficient power, compute, and network bandwidth globally. Token generation capacity—the ability to safely, securely, and efficiently generate AI tokens—is the new metric for global competitiveness and national security.

  2. Context Gap & Machine Data: AI agents without sufficient context make poor decisions (analogous to an ER doctor without patient history). Future AI must be enriched with enterprise data and machine-generated data (time series, sensor readings, logs)—which will comprise 55% of global data growth as agents operate 24/7.

  3. Trust Requires Runtime Governance: AI must be protected from the world (against jailbreaking, prompt injection) and the world must be protected from AI (against rogue behavior). Governance cannot be a static document—it must be implemented dynamically at runtime with observable guardrails.

  4. Shift from Human-in-Loop to AI-in-Every-Loop: Rather than requiring a human in every decision, the paradigm shifts to having AI in every loop while humans remain responsible for high-stakes decisions. This requires fundamental process redesign to accommodate agent-centric workflows.

  5. Physical AI & Multidevice Agents: The mobile ecosystem will be transformed from "device-centric" to "agent-centric," where a single AI agent is accessible across phones, glasses, wearables, and edge devices. This will reshape the entire value chain (OS, app stores, distribution models).

  6. Services Model Not Dead—Evolving: The $300+ billion AI services opportunity lies in six areas: AI engineering (agents), legacy modernization, domain-specific model training, and solution-building—not just raw code generation. System integrators remain essential for validation, verification, and security.

  7. Responsible AI ≠ Regulation ≠ Innovation: Proper governance can drive innovation rather than stifle it. Frameworks must be use-case-specific and evidence-based (as Rwanda demonstrates), not abstract blanket regulations.

  8. India's Unique Position: India's demographic dividend (young, tech-talented population), digital infrastructure (Aadhaar, UPI), massive scale, and development-focused ethos position it to build AI for global problems—not just consume AI built elsewhere. India can be the "use-case capital" and build domain-specific models, not just foundation models.

  9. Human Judgment Irreplaceable in High-Stakes Domains: Military scenario illustrates that AI recommendations with 90% accuracy can fail catastrophically without human oversight. The commander's pause revealed civilian evacuation that algorithms missed—demonstrating why human judgment and moral responsibility cannot be delegated to machines.

  10. Democratization Essential for Impact: Unless AI tools reach MSMEs, blue-collar workers, farmers, and underserved populations, it remains a Fortune 500 story. True transformation requires making AI accessible at the grassroots—via DPI (digital public infrastructure), native language interfaces, and community-based solutions.

Notable Quotes or Statements

"The future is not going to be built by AI alone. The future gets built when humans can confidently put AI to work and delegate jobs and tasks to AI in a way that we feel safe and secure." — Ju Patil, CEO Cisco

"Infrastructure is oxygen for AI. If you don't have enough infrastructure, you're not going to be able to harness the full potential of AI." — Ju Patil, Cisco

"We are building an entirely new global AI industry. The nations that lead this transformation will prosper. Those that merely consume AI built elsewhere will fall behind." — Her Excellency Eva Bush, Deputy PM Sweden

"The future will not be decided necessarily by those that build the biggest models, but rather the ones that build the most trusted systems." — Her Excellency Eva Bush, Sweden

"What is dangerous is not understanding [new technology]. Those who understood the printing press could reach a nation in two weeks and a continent in two months." — Her Excellency Eva Bush, Sweden (on historical analogy)

"AI is not about foundation models only... India will not only be the use-case capital of the world but India will also be the capital of many LLMs that India will build." — Vijay Shekhar Sharma, Founder Paytm

"What does the machine not know? ... A pause revealed something the algorithm could not see. A civilian evacuation had just begun, not yet reflected in the data." — Lt. Gen. Bipul Shinghal, Deputy Chief Army Staff (on meaningful human control)

"AI enabled systems are designed to cause harm. Therefore they must be treated as a weapon and not as software." — Lt. Gen. Bipul Shinghal, Indian Armed Forces

"Nandan [Nilekani] is right—this is not an opportunity gap, it is an execution gap. The real money is in cleaning up trillions of dollars of legacy tech debt." — Salil Parekh, CEO Infosys

"There are some constant in philanthropy that are very important... [We need to ensure] civil society has a voice." — John Palfrey, President MacArthur Foundation

Speakers & Organizations Mentioned

Government & Policy

  • Narendra Modi, Prime Minister of India
  • Her Excellency Eva Bush, Deputy Prime Minister & Minister for Energy and Business, Sweden
  • Her Excellency Paula Angibar, Minister of ICT and Innovation, Rwanda
  • Lt. Gen. Bipul Shinghal, Deputy Chief of Army Staff (Information Systems & Training), India
  • Amitab Kant, former CEO NITI Aayog; moderator

Industry & Technology Leaders

  • Ju Patil, CEO Cisco
  • Cristiano Amon, President & CEO Qualcomm
  • Takahito Tokita, President & CEO Fujitsu
  • Vijay Shekhar Sharma, Founder Paytm
  • Naina Bhasin, Founder Birla AI Labs; Director Aditya Birla Group
  • Sundar Pichai, CEO Google/Alphabet (referenced)
  • Sam Altman, CEO OpenAI (referenced)

IT Services & Enterprise Leaders

  • Salil Parekh, CEO Infosys
  • Kris Gopalakrishnan (referenced as Crispy Vasan/TCS CEO)
  • C. Vijay Kumar, CEO HCL Technologies
  • Arundati Bhattacharya, Chairperson & CEO Salesforce India

Philanthropic & Academic

  • John Palfrey, President MacArthur Foundation
  • Tara Vasarhelyi, Global Head AI & Data Policy, JP Morgan Chase (formerly Obama Admin)
  • Rudra Chadri, VP Observer Research Foundation; moderator

Startups & Entrepreneurs

  • Mana Sharma, CEO Monoi (questioner)
  • Various unnamed startup founders and scale-up leaders mentioned throughout

Companies & Institutions

  • Cisco, Qualcomm, Fujitsu, Google, Microsoft, Amazon, Meta, OpenAI
  • JP Morgan Chase, Salesforce, Infosys, TCS, HCL Technologies
  • Aditya Birla Group, Swatantra Microfinance, Paytm
  • NITI Aayog, Indian Armed Forces, Ministry of IT
  • ASML (Netherlands), ARM (UK), SAP (Germany), Ericsson (Sweden)
  • Indian universities (IIT Bombay, IIT Madras, Delhi University, BITS Pilani)
  • Observer Research Foundation, MacArthur Foundation, Partnership on AI

Technical Concepts & Resources

AI Architectures & Approaches

  • Large Language Models (LLMs): Foundation models trained on public internet data; discussed as baseline but insufficient alone
  • Agentic AI/Agents: Autonomous task-execution systems that understand intentions and make decisions; described as Phase 2 of AI evolution
  • Physical AI: AI deployed on edge devices (robots, sensors, autonomous systems); Phase 3
  • Domain-Specific Models: Specialized models trained on enterprise/sectoral data rather than general-purpose models
  • Structured Foundational Models: Models trained on time-series and tabular data (stock prices, sensor readings, supply chains, weather)—new research vertical at Birla AI Labs
  • Generative AI in Enterprise: Solutions combining LLMs with enterprise data, security, and governance

Infrastructure & Compute

  • Token Generation: The capacity to safely, securely, and efficiently generate AI tokens—new metric for competitiveness
  • Compute Efficiency: Shift from "spiky" inference (chatbots) to steady-state persistent demand (agents)
  • Edge Computing vs. Cloud: Both growing; complementary rather than competitive; seamless distribution across device, edge, network, cloud
  • Quantum Computing: Fujitsu on track for 1,000-qubit machines by March 2027; 250 logical qubits in 2030
  • 6G Networks: Next-generation telecom infrastructure with AI as core component; sensing networks; support for distributed edge AI
  • 2nm & 1.4nm Processors: Next-generation CPUs for data centers (Fujitsu ARM-based, power-efficient)

Data & Context

  • Machine Data: Time-series and sensor data (weather, stock prices, patient vitals, supply chain signals, energy consumption); expected to be 55% of global data growth
  • Enterprise Data Integration: Proprietary data enrichment for competitive differentiation
  • Data Sovereignty: National control over data storage, processing location, and access; "data protection and privacy by design"

Governance & Safety

  • India AI Governance Guidelines: Launched during summit; defines generative AI systems and unintended consequences
  • Delhi Declaration: Framework for responsible AI development
  • Meaningful Human Control: Cannot be delegated to machines; especially critical in defense, healthcare, finance
  • Runtime Governance: Dynamic guardrails injected at runtime (not static policy documents)
  • Black Box → Glass Box: Transparency into data, training methods, decision logic of AI systems

Defense & Security Applications

  • Samadrishti: Battlefield situational awareness software
  • Akashteer & Akasha Shakti: Sensor-shooter fusion systems
  • AI as a Service (AaaS): Indigenously built military applications
  • Autonomous Weapons: Ongoing UN discussions on meaningful human control and accountability

Tools & Platforms Referenced

  • Cisco Topaz Fabric (Infosys): Integration layer for multiple foundation models and agents
  • Birla AI Labs Research Platform: AI-native research and productivity tool combining agentic search, real-time data, multimodal intelligence
  • Salesforce Einstein: Enterprise AI platform with governance, auditability, observability
  • HCL's Custom Silicon: 2nm chips for enterprise partners

Research & Academic Work

  • Time² (Time to Time) Paper: Research on whether time-series models understand market dynamics or just fit curves; presented at Oxford AI Summit and World Summit AI 2025
  • Delhi University Study (Birla AI Labs): Measuring impact of LLM usage on curiosity and cognitive agency; results to be presented at King's College June 2025
  • Curriculum Development: TCS, Infosys, HCL working with Ministry of IT on AI curriculum for universities

Methodologies & Frameworks

  • Use-Case-by-Use-Case Risk Assessment: Rwanda's approach—evaluate and regulate specific applications rather than blanket policies
  • Responsible AI Frameworks: Industry-wide standards (mentioned by multiple speakers) that integrate ethics, transparency, and accountability
  • Agile Regulatory Posture: Evidence-based, adaptive regulations that evolve with deployment experience
  • Outcome-Based Contracting: Shifting from input-based (time & materials) to output-based pricing as AI systems mature

End of Summary