India AI Impact Summit 2026

mahaai-building-safe-secure-and-smart-governance

8 min read

TRACK:Safe and Trusted AI
SPEAKERS
  • Davinder Sandhu, Primus Partners
  • Dr. Amit Kapoor, Institute for Competitiveness
  • Dr. Anupam Chattopadhyay, Nanyang Technological University
  • Dr. Ganesh Ramakrishnan, IIT Bombay
  • Mr. Ashish Tewari, Infosys Responsible AI Office (India)
  • Mr. Bhuvnesh Kumar, UIDAI
  • Mr. Devroop Dhar, Primus Partners
  • Mr. Ranjeet Goswami, Tata Consultancy Services
  • Mr. Suresh Sethi, Protean eGov Technologies
  • Mr. Vikas Rastogi, Department of Agriculture
  • Mr. Virendra Singh, Government of Maharashtra
  • Ms. Beena Sarkar, ServiceNow
SESSION OVERVIEW

This session will showcase Maharashtra's approach to building safe, secure and citizen-centric AI governance. It will examine how the State is embedding ethics, data protection, security and AI system for public services, safety and administration. The discussion will outline Maharashtra's policy roadmap for responsible AI adoption and highlight how AI can drive economic growth and social good while remaining transparent, accountable and trusted.

VIDEO RECORDING

MahaAI: Building Safe, Secure and Smart Governance

Detailed Summary

1. Introduction & Setting the Stage

  • Opening remarks by the moderator thanked the audience and introduced the theme: AI is already reshaping governance, markets, public services, and geopolitics. The central question posed was not whether AI will shape governance, but how governance will shape AI.
  • A concise definition of the “governance paradox” was given: over‑regulation stifles innovation; under‑regulation risks harm. The moderator argued that “intelligent governance” – human‑centred, transparent, risk‑based, and globally coordinated – is the answer.

2. Keynote Address – “Maha AI: A Living Laboratory”

Speaker: Ashish Shailar (Hon’ble Minister of IT & Cultural Affairs)

  • Context: The AI Impact Summit 2026 – the first of its global series hosted in the Global South – bringing together >20 heads of state, 60 ministers, and hundreds of AI leaders.
  • Maharashtra’s Vision: Under Chief Minister Devendra Fadnavis, the state positions itself as a living laboratory for AI‑driven governance.
  • Flagship Projects
    • Maha Crime OS – AI‑enabled crime‑prevention platform (demoed by Microsoft’s Satya Nadella). Faster investigations, transparent processes.
    • Maha IT – “intelligent government infrastructure”: a cloud‑native, modular, API‑driven backbone that powers smart recruitment, AI‑based property mapping, real‑time urban dashboards (traffic, weather, civic issues), flood‑management and smart‑mobility pilots.
  • Five Pillars of a Safe‑Secure‑Smart Governance Stack
    1. Compute & Cloud at Scale
    2. High‑Quality Public Data Sets
    3. State AI Governance Framework
    4. Inter‑operability & Standards
    5. Capacity‑Building & Skill Development
  • Policy Perspective: AI must humanise the state, not distance it from citizens. The goal is “scale apathy through insight”.
  • Digital Health & Disinformation: Emphasised the need for robust cybersecurity, digital‑literacy, hybrid verification ecosystems, and a “digital‑health” policy akin to physical health.

Announcement: The government called on global technology partners to co‑create with Maharashtra, leveraging world‑class AI platforms while ensuring sovereignty.

3. Panel Transition – Introducing the Panelists

The moderator invited the panelists onto the stage:

  • Suresh Sethi (Protean eGov)
  • Ranjeet Goswami (TCS)
  • Beena Sarkar (ServiceNow / Women for Ethical AI)
  • Devroop Dhar (Primus Partners)
  • (Other panelists appeared briefly; names were occasionally garbled in the transcript.)

4. Economic Impact & Data Monetisation

Speaker: Devroop Dhar (Primus Partners)

  • Job‑Market Shifts – Citing a NITI‑Aayog analysis: from the 1950s‑2020 era, post‑graduates & engineers held a 95 % employment probability. Since the 2020s, the growth of physical‑skill jobs (masons, bricklayers, home‑carers) is now 0.65 %, indicating AI’s disproportionate impact on high‑skill employment.
  • Data as a Public Asset – The state data authority is working to ‘cash‑in’ public data at scale while preventing foreign exploitation. Two illustrative sectors:
    1. Pharmaceuticals – Massive health‑data sets (population health, disease incidence) that can be monetised for domestic R&D.
    2. Government Orders (GRs) – Over 150,000 complex orders exist; the state is partnering with IIT‑Bombay (Prof. Ganesh Ramakrishnan) to build a small‑language‑model (MahaGPT) that parses orders, extracts the latest legal stance, and replies to both officials and citizens.

Key Insight: A single AI layer (MahaGPT) will serve both bureaucrats and citizens, turning a “maze of orders” into a searchable knowledge base.

5. Cybersecurity, AI & Quantum‑Computing Threats

Speaker: Bhuvnesh Kumar (UIDAI) – “Cyber” segment

  • Maharashtra Cyber‑Security Project – Launched five years ago under the Chief Minister’s direction.
    • AI‑Powered Tool‑chain – Dark‑web monitoring, threat analysis, social‑media scanning, ransomware detection, cyber‑bullying mitigation.
    • One‑Stop Helpline (1930) – Over 150 cyber‑consultants field reports via a single number.
  • Impact (first 6 months):
    • ₹1,000 crore frozen from fraudsters and returned to victims.
    • ≈70 young women rescued from cyber‑bullying, blackmail, and potential suicide thanks to AI‑driven tracking.
  • Echos of Pehelgaam Report – During a conventional India‑Pakistan clash, >1 million nation‑state cyber‑attacks were launched (APT groups from Indonesia, Pakistan, Turkey). AI‑based threat‑intel tools (Luminar, Cognite, Pathfinder) thwarted many attacks.
  • Quantum‑Computing Warning – Quantum computers (hundreds of millions of qubits) could break RSA, blockchain, and banking encryptions in seconds. Current national investment ₹1 billion versus $15–20 billion by China. Maharashtra needs a quantum‑readiness strategy to safeguard financial and civic infrastructure.

Recommendation: Simultaneous strengthening of cyber‑defence AI and early investment in quantum‑safe cryptography.

6. Research Frontiers – Deep‑Fake Detection

Speaker: Dr. Anupam Chattopadhyay (Nanyang Technological University)

  • Spin‑off Company – Focus on AI for detecting deep‑fakes (audio, video, images).
  • Data‑Scarcity Challenge – Lack of labeled Marathi/Hindi deep‑fake corpora. The team scrapes internet data and builds a synthetic‑data pipeline while ensuring ground‑truth verification via cross‑referencing with news reports.
  • Noise Robustness Experiments – Clean samples yielded high detection rates; adding synthetic ambient noise degraded performance. Training with noisy augmented data restored accuracy.
  • Privacy‑Preserving Techniques –
    • Fully Homomorphic Encryption (FHE) – Feasible but computationally heavy.
    • Federated Learning with Differential Privacy – Allows multiple parties to train a shared model without exposing raw data or proprietary weights.
  • Deployment Constraints – Participated in a UK Home Ministry hackathon that required air‑gap execution (no cloud connectivity). Required model compression, mixture‑of‑experts, and edge‑device optimisation – a research‑to‑product pathway still being refined.

Takeaway: Real‑world deep‑fake defence needs robust data pipelines, noise‑tolerant models, and privacy‑preserving distributed training.

7. Digital Public Infrastructure (DPI) & AI Layer

Speaker: Suresh Sethi (Protean eGov Technologies)

  • Population‑Scale DPI – Identity (UID), payments (UPI), document storage (DigiLocker) already in place. This creates a foundation for AI‑enabled eligibility and subsidy delivery.
  • Dynamic Eligibility vs. Static Identity – Machine‑readable “verifiable credentials” (blue‑dot attributes) enable AI to match citizens to appropriate benefits in real time.
  • Predictive Governance – AI can foresee income distress signals and pre‑emptively trigger subsidies.
  • Error Types:
    • Inclusion Error (Leakage): Benefits go to ineligible persons.
    • Exclusion Error (Denial): Eligible persons miss out.
  • Guardrails Required:
    • Explainability – Every AI decision (grant or denial) must be transparently justified.
    • Auditability – Immutable logs for post‑hoc review.
    • Human Redressal – A clear escalation pathway for citizens to contest AI outcomes.

Implication: Embedding AI into DPI can transform service precision, proactivity, and trust, provided strong governance safeguards are baked in.

8. Role of Large Tech Companies – Collaborative Blueprint

Speaker: Ranjeet Goswami (TCS)

  • Purpose‑First Lens – AI should advance welfare & happiness (the summit’s tagline), not merely efficiency.
  • Holistic Integration – Government departments should share a unified citizen database (e.g., the Aadhaar ecosystem) to avoid siloed data.
  • Step‑wise Approach:
    1. Common Data Layer – Connect every department to a single source of truth (the ADHA database).
    2. Platform Intelligence – Infuse AI services on top of this layer to enable cross‑departmental insights.
    3. Iterative Pilots – Begin with low‑risk use‑cases (smart recruitment, traffic analytics) before scaling.

Key Message: Technology partners must align their product roadmaps with public‑good outcomes, ensuring that AI deployments are inclusive and citizen‑centric.

9. Ethical AI & Gender‑Bias Considerations

Speaker: Beena Sarkar (ServiceNow / Women for Ethical AI – South Asia chapter)

  • Problem Framing: Many AI products focus on hardware (e.g., smart glasses) without considering societal impact. Past failures (Google Glass) illustrate privacy‑violation risks.
  • Safety‑First Framework (India Safety Institute – 2025): Any new device must pass a first‑line safety assessment evaluating:
    • Potential for non‑consensual imaging
    • Threat to women & children
    • Possibility of amplifying gender‑based harassment
  • Ethical Evaluation Model – “Kali vs. Rakta Bija” – A metaphor for harmful vs. beneficial technology; decisions should err on the side of preventing systemic harm.
  • Policy Recommendation: The Institute should vet new AI‑enabled hardware before market entry, akin to firearm regulation, to prevent “50 % of the population” from being exposed to unsafe technology.

Takeaway: Ethical AI governance must explicitly address gender‑based risks and include a pre‑market safety clearance mechanism.

10. AI for Tier‑2/3 Cities – Scaling the Benefits

Speaker: Dr. Amit Kapoor (Institute for Competitiveness)

  • Skill‑Gap Reality: Only ~20 % of Maharashtra’s 9 crore workforce resides at skill levels 3‑4; 80 % are at levels 1‑2. Upskilling is critical for AI adoption.
  • Connectivity Bottleneck: Average broadband speed in Mumbai ≈ 58 Mbps; many tier‑2/3 towns lag far behind, hampering AI‑driven services. Calls for war‑footing infrastructure upgrades.
  • Talent & Infrastructure Concentration: Pune houses ≈ 16 % of India’s tech workforce—an asset for building state data‑centers, AI labs, and R&D hubs.
  • Sectoral Opportunities:
    • Nutrition Monitoring: AI can map malnutrition at pin‑code granularity (currently ≈ 50 % of Maharashtra’s population is malnourished).
    • Water & Sanitation: Predictive analytics for water‑quality and waste‑management.
    • Education: AI‑enabled personalized learning platforms for tier‑2/3 schools.
  • Societal Risks: AI could become a “dumping ground” for low‑quality content, deepening digital addiction (e.g., “doom‑scrolling”). Emphasised the need for digital‑wellness curricula.
  • Under‑Employment Challenge: Nearly 50 % of the state’s workforce is under‑employed; AI‑driven reskilling programmes are essential.

Conclusion: If Maharashtra simultaneously upgrades connectivity, scales skill development, and embeds AI responsibly, it can set a replicable model for the rest of India.

11. Closing Remarks & Photo Session

  • The moderator thanked all panelists, highlighted the collective commitment to “govern intelligence with wisdom”, and invited the panel for a group photograph.
  • Virendra Singh (senior government official) joined the panel for the photo, signalling continued high‑level support.

Key Takeaways

  • Intelligent governance – Balances rapid AI innovation with risk‑based, globally coordinated policy.
  • Maha AI’s five‑pillar stack (compute, data, governance, standards, capacity) guides Maharashtra’s AI‑first agenda.
  • Maha Crime OS and Maha GPT illustrate real‑world AI applications that improve public safety and bureaucratic efficiency.
  • Economic displacement is evident: high‑skill jobs are shrinking relative to physical‑skill occupations; proactive data‑monetisation strategies are required to capture public‑good value.
  • Cyber‑security gains – AI‑driven helpline (1930) froze ₹1,000 crore and saved ≈ 70 lives in six months; quantum‑computing threats demand early mitigation.
  • Deep‑fake defence relies on diverse, labeled datasets, noise‑robust training, and privacy‑preserving federated learning.
  • Digital Public Infrastructure (DPI) is a prerequisite for AI‑enabled eligibility, predictive welfare, and auditability; explainability and human redress are non‑negotiable.
  • Tech‑industry collaboration must be purpose‑driven, emphasizing a single citizen‑centric data layer and incremental pilots.
  • Ethical AI – Gender‑bias and privacy concerns require a dedicated safety institute to vet hardware and algorithms before market entry.
  • Tier‑2/3 uplift – Closing the skill gap, expanding broadband, and leveraging Pune’s tech talent are essential for statewide AI diffusion.
  • Governance outcome: Maharashtra aims to become a global benchmark for safe, secure, and smart AI governance, positioning India as a leader in responsible AI deployment.

See Also:

Graph View

Backlinks