India AI Impact Summit Insights
All sessions

AI, Policy, and the Rule of Law: A Policymakers’ Dialogue

Contents

Key Takeaways

Session 1: Governance & Rule of Law

  1. Rule of law and democratic accountability are prerequisites for AI, not obstacles to innovation—trust enables adoption, predictability encourages investment, accountability sustains legitimacy.

  2. Inclusive design from inception, not retrofit inclusion—AI tools designed by minorities for minorities will replicate and amplify existing power imbalances; diverse representation in development is non-negotiable.

  3. Different regulatory approaches (EU risk-based, India techno-legal) can be complementary, not competing—convergence around human dignity, democratic accountability, and equitable access creates global interoperability without erasing local context.

Session 2: Financial Services & Fraud Detection

  1. Move from "rule-based blocking" to "behavioral anomaly detection"—AI enables real-time network intelligence across institutions, catching sophisticated fraud while reducing false positives that exclude legitimate users.

  2. Explainability is not optional in regulated finance—the JCT framework (Justifiable, Contestable, Traceable) ensures AI systems can be audited by customers, auditors, and regulators; governance is the foundation of long-term value.

Session 3: Energy & Grid Modernization

  1. Demand flexibility, not just demand response—AI enables predictive load shifting (water pumping, EV charging, thermal storage) to match variable renewable generation, reducing need for new grid capacity investment.

  2. Platform thinking over point solutions—India Energy Stack model of interoperable building blocks allows 84 heterogeneous discoms to contextualize solutions without proliferating incompatible systems.

Session 1: AI, Policy, and the Rule of Law: A Policymakers' Dialogue

Executive Summary

This policymakers' panel discussed how India and Germany are approaching AI governance through different regulatory models—India favoring a constitution-based, techno-legal approach with soft-touch regulation, while the EU uses risk-based legislation. The consensus emphasized that rule of law, democratic accountability, and inclusive design are essential to preventing power concentration and ensuring AI benefits all socioeconomic levels.

Key Topics Covered

  • India's constitution-based AI governance approach vs. the EU's risk-based AI Act
  • Balancing innovation with responsible regulation
  • Inclusivity and accessibility as core governance principles
  • Power concentration risks in AI development
  • Cross-jurisdictional learning and regulatory harmonization
  • Democratic resilience and citizen protection
  • Role of courts, legislatures, and regulators in AI oversight
  • Data governance and digital public infrastructure (DPI)
  • Social and political sensitivities in diverse democracies

Key Points & Insights

  • India's Approach: India is pursuing a "soft touch" regulatory model based on constitutional principles (equality, transparency, accountability, inclusivity) rather than hard statutory rules, recognizing the country's digital divide and rural-urban data disparities.

  • Techno-Legal Model: Before enacting new AI legislation, India evaluates existing laws (privacy, consumer protection) and embeds controls into system design rather than creating separate AI-specific laws.

  • Power and Rules: As Ray Dalio noted (cited by speaker), rules are typically demanded by the weak to control the powerful—making rule-based governance critical when powerful tech actors dominate AI development.

  • Inclusivity as Design Principle: AI governance cannot simply mean giving everyone tools; it requires inclusive design from inception, ensuring diverse voices, languages, ethical values, and communities are represented in tool development.

  • JCT Framework (from second panel): Justifiability, Contestability, and Traceability are essential for governance—decisions must be explainable, contestable by affected parties, and traceable to identify biases.

  • Regulatory Learning: Policymakers must adopt a "humble approach," gathering weak signals early, including stakeholders throughout the policy cycle, piloting before scaling, and maintaining feedback loops to learn while doing.

  • Court-Led Innovation Path: India historically has allowed courts to develop solutions (via common law and judicial interpretation) before legislation—enabling adaptive, problem-driven governance rather than static rules.

  • Citizen-Centric Risks: AI systems linking voter behavior, government benefits, and social media profiles pose existential threats to democratic freedoms and constitutional protections in diverse societies.

Session 2: Harnessing AI for Fraud Prevention and Financial Inclusion (India-Singapore Business Forum)

Executive Summary

This panel examined how AI can transform financial fraud detection and inclusion in India's high-velocity digital payment systems, moving from rule-based blocking to intelligent network analysis. Speakers emphasized that explainability, regulatory accountability, and human-in-loop oversight are non-negotiable for scaled AI deployment in regulated finance.

Key Topics Covered

  • AI-driven fraud detection vs. legacy rule-based systems
  • Real-time risk assessment across transactions and networks
  • Mule account networks and cross-institutional anomaly detection
  • False positives and customer exclusion trade-offs
  • Explainability, contestability, and traceability (JCT framework)
  • Regulatory approval workflows and human oversight
  • AI as capacity-building and capability enabler
  • Onboarding speed and loan approval timelines
  • India-Singapore partnership in fintech governance

Key Points & Insights

  • Scale and Velocity Problem: UPI processes 20+ billion transactions monthly with projected fraud losses exceeding 1 lakh crores—rule-based thresholds cannot detect sophisticated fraud at this scale or speed.

  • Network Intelligence: AI identifies behavioral anomalies across multiple accounts and institutions simultaneously, detecting mule networks moving money in milliseconds across branches—something single-institution rule engines cannot do.

  • False Positives Create Real Exclusion: Blocking transactions above certain values catches fraud but also excludes legitimate users, especially first-time borrowers and MSMEs whose livelihoods depend on those transactions.

  • AI as Capacity, Not Just Capability: AI initially provides operational capacity (doing in 2 minutes what took 2 hours) before adding new capabilities; this "capacity first" mindset helps leaders accept AI systems even when performance is merely on-par with human performance.

  • JCT Governance Framework:

    • Justifiable: Any decision (credit denial, transaction block) must be justified
    • Contestable: Affected parties must have recourse to challenge decisions
    • Traceable: Decision pathways, biases, and model versions must be auditable

  • Stakeholder Accountability: Three critical stakeholders demand different governance: customers (transparency and recourse), auditors (traceability and evidence), and regulators (human-in-loop oversight and accountability).

  • Regulatory Requirement for Human-in-Loop: Video KYC in India required human verification even with automated checks—regulators mandate human oversight to ensure real-person onboarding before AI can make autonomous decisions.

  • Governance as Value Creation: Governance should not be viewed as compliance overhead but as a pathway to building trustworthy, resilient products that generate long-term value and innovation.

Session 3: AI for Power—Accelerating the Clean Energy Transition

Executive Summary

This panel addressed how AI and advanced analytics are essential for managing increasingly complex electricity grids with high renewable penetration, distributed energy resources, and new flexible loads like data centers and EVs. The discussion emphasized the need for system-level coordination through open digital platforms (India Energy Stack model) rather than siloed solutions.

Key Topics Covered

  • AI forecasting for variable renewable energy generation
  • Demand flexibility and demand response optimization
  • Smart building management and thermal energy storage
  • Electric vehicle charging infrastructure and flexibility
  • Data center flexibility as a grid asset (not just a load)
  • Predictive asset health monitoring and maintenance
  • Regulatory frameworks and performance-based incentives
  • Capacity building for grid operators and regulators
  • South-South learning and global South grid modernization
  • Digital twins and advanced analytics for utilities
  • Societal platforms and India Energy Stack principles

Key Points & Insights

  • Growing Grid Complexity: Three mega-trends drive need for AI: (1) increasing electrification of end-uses, (2) rising variable renewable penetration (from 2-3% a decade ago to 30% by 2030), and (3) long-term net-zero targets with short-term implications.

  • Forecasting Critical for India: Hourly, location-specific solar and wind forecasting is essential—not just global weather accuracy but minute-by-minute cloud cover prediction at individual panel locations.

  • Demand Response Evolving to Demand Flexibility: The model is shifting from grid operator triggering load reduction to proactive, predictive participation where consumers (and now data centers, EV chargers, and water pumping systems) contribute to grid balancing.

  • Data Centers as Grid Assets: Properly sited and sized data centers with 20-30% demand flexibility can support grid balancing during peak solar generation and serve as firm flexible loads—not pure burdens if planned with renewables alignment.

  • Hourly Renewable Matching Target: Major tech companies have moved beyond 100% yearly renewable matching to 100% hourly matching (e.g., Google, Nvidia targets for 2030-2035), requiring real-time coordination with grid operators.

  • AI Operational Value in Delhi: BRPL demonstrated 9-year AI deployment in demand forecasting (spot/day-ahead/intraday), achieving 70-80% cost optimization for utilities; currently deploying predictive asset health monitoring on all 23 KV and 11 KV feeders.

  • Regulatory Sandboxes and Performance Incentives: CERC (Central Electricity Regulatory Commission) enables P2P blockchain-based trading pilots; utilities adopting digital solutions need regulatory incentives for REC curtailment reduction, forecasting accuracy, and outage reduction.

  • Skills Gap Crisis: Energy sector faces 2x wage gap vs. tech sector for similar AI/data roles—talent gravitates to tech, starving energy sector of expertise needed for grid digitalization.

  • Common Building Blocks Over Siloed Solutions: India Energy Stack approach identifies 5 core use cases (55 variations) applicable to all 84 discoms rather than requiring 15+ separate RFPs per discom for different digital solutions.

Notable Quotes or Statements

On Power and Rules (Dr. Patra, MP):

"Rules are demanded by the weak to control the powerful, and the powerful don't operate by any rules. They change the rules once they leave power."

On Inclusive Design (Ivana, Wipro):

"Democratizing AI doesn't simply mean giving everybody tools. It means that you are inclusive in the way that these tools are designed in the first place. We can't have tools designed by a small minority."

On Governance as Value (Manish, 811 Coins):

"We have to stop seeing governance of artificial intelligence as compliance. We have to see governance as the way to create long-term value, innovation, growth, productivity."

On AI and Capacity vs. Capability (Suresh, Protego):

"AI is first catalyzing capacity and then adding new capabilities. What took two hours can now be done in two minutes."

On the Grid Transformation (Sujit, IIA):

"A grid is no longer a monolith or a unidirectional thing. It's already a bidirectional, multi-directional thing."

On Regulatory Humility (Dr. Marcus, TUM):

"We don't know where the field is moving at the moment. So we need to have a more humble approach and think about what mechanisms can we put in place to learn while doing."

On Future Energy Livelihoods (Sujit, IIA):

"Maybe somebody in the village will start thinking: should I buy a cow for livelihood or get a non-self-used energy solar rooftop and battery to generate income?"

Speakers & Organizations Mentioned

Government & Policy

  • Dr. Patra (MP, Member of Parliamentary Standing Committee on Communications and Information Technology; represents Andhra Pradesh)
  • Stephan Stampy (CAT, Rule of Law Program in Asia, Singapore)
  • Christina Cinemos (Minister of State for Digitalization and Innovation, Hessen, Germany)
  • Mr. Singhi (Additional Secretary, Ministry of External Affairs, India)

Financial Services & Fraud Prevention

  • Suresh (Managing Director & CEO, Protego Technologies)
  • Sarab Mittal (Country Head, Strategy Transformation Analytics, DBS India)
  • Bhuvan Noda (Chief Executive Officer, AI Division, Mahindra & Mahindra)
  • Sujay Gosh (Managing Director, TMS)
  • Manish Aagarwal (President, Business, 811 Coins/Mahindra Bank)
  • Niraj Agarwal (Managing Director & Senior Partner, Boston Consulting Group Singapore; Moderator)

Energy & Grid

  • Dr. Mahesh Patanka (MPEN Systems)
  • Sadhhat Singh (International Energy Agency, leading Energy & AI work program)
  • Namita Mukharji (Strategic Planning Specialist, International Solar Alliance)
  • Sujit Naya (DPI/Energy Stack) / Sujit (Infrastructure/DPI work)
  • Abishek Ranjan (Leading energy innovation, Delhi grid digitalization)
  • Rishi Nalin (Climate Collective, ElectronVibe program)

Academic & Think Tanks

  • Dr. Marcus Zebat (Technical University Munich, TUM Think Tank)
  • Ivana (Wipro, AI Governance)

Organizations

  • FIKI (Confederation of Indian Industry)
  • India-Singapore Business Round Table (ISBR)
  • International Solar Alliance (ISA)
  • International Energy Agency (IEA)
  • Climate Collective & ElectronVibe
  • NPCI (National Payments Corporation of India)
  • CERC (Central Electricity Regulatory Commission)
  • DBS India
  • Wipro
  • Boston Consulting Group

Technical Concepts & Resources

AI & Governance Frameworks

  • JCT Framework: Justifiability, Contestability, Traceability (for financial services and regulated sectors)
  • Constitution-based AI governance (India's approach, emphasizing equality, transparency, accountability, inclusivity)
  • Risk-based regulation (EU AI Act model)
  • Techno-legal approach: Evaluating existing sectoral laws before new AI legislation
  • Regulatory sandboxes: CERC P2P blockchain trading pilots; UPRC experiments

Financial Services Technologies

  • UPI (Unified Payments Interface): 20+ billion transactions/month in India
  • Video KYC (Know Your Customer): AI-assisted biometric verification with mandatory human oversight
  • Mule account networks: Sophisticated fraud pattern where money moves through multiple accounts in milliseconds
  • AI-powered anomaly detection: Network-level behavioral analysis across institutions
  • Deviation Settlement Mechanism (DSM): 15-minute interval basis for renewable energy market participation

Energy & Grid Technologies

  • Demand forecasting: Spot market, day-ahead, and intraday (9-year deployment at BRPL achieving 70-80% cost optimization)
  • Predictive asset health monitoring: Drone/satellite imagery for feeders to enable predictive maintenance
  • Demand response → Demand flexibility: Proactive load shifting for renewable matching
  • Thermal energy storage systems: Pre-cooling buildings to shift cooling load to solar peak hours
  • Vehicle-to-Grid (V2G) and Vehicle-to-Load (V2L): EV battery as flexible grid asset
  • Water pumping optimization: Seasonal storage opportunity in Uttarakhand pilot (40% of state load is water pumping)
  • Digital twins: Utility-scale models for planning, operation, and resilience testing
  • Smart inverters: Distributed energy resource management at device level
  • Data center flexibility: 20-30% demand shaving potential without compromising compute capability

Data & Infrastructure

  • Digital Public Infrastructure (DPI): India's model (Aadhaar, UPI, DigiLocker)
  • India Energy Stack: Proposed interoperable standards for grid modernization
  • Societal Platform / Societal Thinking: Design principles for population-scale coordination
  • Smart meter rollout: 250 million smart meter deployment program (India)
  • RDSS (Real-time Data & Supervisory Systems): Base layer for utility digitalization
  • Open data platforms: For renewable forecasting (multi-stakeholder access to improve AI models)

Regulatory & Metrics

  • REC (Renewable Energy Certificate) curtailment reduction
  • Forecasting accuracy metrics: Variable renewable energy (hourly, location-specific)
  • Outage reduction KPIs
  • Performance-based incentives for utilities adopting digital solutions
  • Grid code updates: Continuous revision to reflect decentralized, bidirectional grid realities

Global Context

  • 148 bills on AI from 32 countries (in recent years; three broad legislative approaches: no-change, risk-based, rights-based)
  • EU AI Act: Risk-based classification and unacceptable risk standard (strict liability)
  • US Right to Publicity model: Rights-based approach to AI
  • India's 20-30 bills/acts per year vs. EU 300+: Legislative pace differences affecting regulatory strategy

Note: This transcript spans three distinct panel sessions at the India AI Impact Summit.