The AI Openness Forum: Building Trust Through Transparency

Contents

Executive Summary

This multi-session AI summit transcript covers three major conversations: (1) cross-border AI collaboration and sovereignty, (2) AI assurance and standards implementation, and (3) generative AI applications in agriculture. The overarching theme emphasizes that effective AI deployment requires not just technical innovation, but alignment across policy, trust mechanisms, data infrastructure, and ecosystem partnerships—particularly for emerging economies like India to capture value and maintain agency in AI-driven systems.

Key Takeaways

Sovereignty is about control over critical decision layers, not isolation. Openness and modularity strengthen sovereignty by preventing single-system dependencies.
Trust precedes all cross-border AI collaboration. Without trust mechanisms (shared testing, interoperability standards, joint validation), misaligned AI systems create systemic risk.
Small, context-aware AI models deployed at the edge outperform expensive LLMs for agriculture and critical infrastructure. DPI and frugal engineering are India's competitive advantage.
Certification and assurance are now market-driven, not regulatory theater. Organizations pursuing ISO 42001 are doing so because customers demand proof of governance—this is real incentive, not checkbox compliance.
Agricultural transformation requires structural reforms (policy alignment, incentive design, inclusive governance) before technology scaling. AI is an accelerant for good systems, not a fix for broken ones.

Key Topics Covered

Cross-Border AI Collaboration & Sovereignty

AI as a coordination layer for critical infrastructure and systems
Agency vs. access: enabling local capacity to shape AI systems
Managing different regulatory environments and AI ecosystems
Trust as a prerequisite for international AI partnerships
Ecosystem design: modularity, openness, and fair value distribution

AI Assurance & Standards

ISO 42001 certification and management system standards
Translating ethical principles into quantifiable, testable governance
Risk-based assurance frameworks (financial services as a model)
Sector-specific maturity levels (healthcare, finance, defense leading)
Agentic AI risks: privacy exfiltration, trust gaps, communication protocols
The role of testing, benchmarking, and continuous monitoring

Generative AI in Agriculture

Precision farming and climate adaptation using AI models
Post-harvest value chain optimization and food security
Digital public infrastructure (DPI) as an enabler
Farmer producer organizations (FPOs) as aggregation points
Contextual vs. data-driven approaches; small, focused models vs. LLMs
Gender-inclusive adoption; structural reforms needed before technology scales
Sovereign farmer identity systems and unified data infrastructure

Key Points & Insights

AI as Coordination Infrastructure: AI is not just about smart models—it's becoming the coordination layer for critical systems (energy, agriculture, logistics, health). This shifts the question from "is this AI good?" to "who gets to shape how systems interact?" and "how do misaligned AI ecosystems create systemic risk?"
Agency Over Access: Giving countries/communities mere access to AI is insufficient. They need agency—the capacity to adapt systems locally, question outputs, adjust behavior in context, and shape decisions. Without agency, nations are "coordinated by systems designed elsewhere."
Context-Driven > Data-Driven: In agriculture and sensitive domains, shifting from purely data-driven to context-driven AI architectures reduces computational overhead, improves relevance, and enables deployment on edge devices. A few well-chosen parameters can replace thousands of data points.
Ecosystem Design Requires Trust Infrastructure: Successful cross-border AI collaboration demands shared mechanisms: testing environments, interoperability standards, cross-border validation, mutual safety principles. Global coordination does not require one authority; it requires alignment without centralization.
ISO 42001 as Market Driver (Not Regulatory Mandate): Certification against ISO 42001 is being driven by supply chain demand, not regulation. India is #2 globally in accredited ISO 42001 certifications (Axis Bank certified first globally), suggesting market pull for assurance is real.
Structural Issues Before Technology: In agriculture, data collection, policy alignment, public-private partnerships, and separation of welfare from business must be resolved before AI scaling occurs. Technology alone cannot solve governance, subsidy, or incentive misalignments.
Frugality as a Feature: High-performing startups and deployments in India succeed through extreme frugality, not capital lavishness. Edge computing, small models, and DPI-based infrastructure are more viable than expensive cloud-based LLMs.
Women Farmers as Neglected Stakeholders: Women constitute a large and growing portion of agricultural labor but are often excluded from schemes, training, and credit access. Gender-inclusive design is essential for equitable AI deployment.
Data Infrastructure as Public Good: Shared, open, democratized data infrastructure (not data itself) drives innovation at scale. UPI's success demonstrates how API-based infrastructure (not data monopolies) enables ecosystem growth and cost reduction.
Climate and Ecological Systems Require Strong AI: Addressing climate resilience and ecosystem health in agriculture requires "strong AI architectures" that integrate multiple domains (soil, water, atmosphere, landscape) seamlessly—not black-box solutions. Local complexity and trigger parameters must be understood.

Notable Quotes or Statements

On AI as Coordination Layer:

"AI is becoming a coordination layer for society. It's starting to help complex systems adjust to each other... Once coordination becomes automated, the question is no longer just about performance. It becomes about who can participate, who can shape it, and how safely these systems work together." — Alexandra Borcoff, CINTF

On Agency vs. Access:

"It only works when all participants have the capacity to make their own independent choices and shape their own life and future. We call it agency, not just access... Countries don't take part in the coordination. They are simply coordinated by systems designed elsewhere." — Alexandra Borcoff

On Ecosystem Trust:

"Ecosystem is a trust-based gathering of everybody interested in the topic. Value is distributed transparently and fairly... It's not me, it's us. Let's do this together." — Basil (Nokia)

On Frugal Assurance:

"Assurance at this point is not just a nice to have—it is demanded in procurement processes and it's a way of building trust across the supply chain." — Natasha Crampton, Microsoft

On Agriculture AI Challenges:

"Hope is not a bad strategy, but it should not be the only strategy. We should be excited about [AI] but the structural issues... need to be sorted out first." — Anand Chandra, Arya

On Data Infrastructure as Public Good:

"Data is essential, but access to data and democratization is even more essential... [Through] shared infrastructure, you have more API economy, you bring the cost down." — Malik Bansali, Neteb Software

On Climate-Aware AI:

"We cannot have a black-box understanding... Strong AI architectures will have a very strong role to play... The entire ecological dynamics should be put on one scale seamlessly." — Aloki Mukhaji, Lead Connect / BL Agro

On Policy Enablement:

"We want these models to be sovereign at the local setting... Ultimately, control of patient data goes to the person. That's what ABDM [Ayushman Bharat Digital Mission] is about." — Dr. Mansukh Mandaviya (implied, healthcare context)

Speakers & Organizations Mentioned

Policy & Government

Sanjay Kumar Agarwal – Joint Secretary, Department of Agriculture & Farmer Welfare, Government of India
Dr. Mansukh Mandaviya (implied) – Healthcare/ABDM context
Shri Sanjay G – Government agriculture leadership

Research & Academic Institutions

Alexandra Borcoff – CINTF (Centrum för Teknologi och Samhälle), leading collaborative AI research in Europe
Carson Maple – Alan Turing Institute (UK), Director of Cyber Security & AI Assurance
Caster Maple – University of Warwick, National Hub for Edge AI
Raj Patel – VP AI Transformation, Holistic AI (UCL spinout)

Industry & Technology

Natasha Crampton – VP & Chief Responsible AI Officer, Microsoft
Tim Maggar – BSI (British Standards Institution), AI Certification & Testing
Basil / Pari – Senior VP Strategic Government & Industry, Nokia
Anil Sharma – Global Head, TCS Co-Innovation Network
Anand Chandra – Co-founder, Executive Director, Arya
Malik Bansali – CEO, Neteb Software
Simon Torson Vibush – Chairman FIKI, CEO Bayer Crop Science (India/Bangladesh/Sri Lanka)
Dr. Aloki Mukhaji – Director Research & Analytics, Lead Connect; Chief System Scientist, BL Agro Industries

International & Civil Society

Elizabeth For – WFP (World Food Program), Country Director India
Sager – Research Manager, CINTF, Norway (World Cafe moderator)

Organizations Mentioned

Microsoft – M365 Copilot, responsible AI framework
CINTF – European research institute; runs EU project OpenMod for Africa
Holistic AI – AI governance SaaS; testing for agentic AI, explainability
BSI – British Standards Institution; ISO 42001 certification body
TCS – Tata Consultancy Services; co-innovation networks
Nokia – Telco/tech; promoting open interfaces, modular ecosystems
Applied Ventures / Applied Materials – Semiconductor, venture investing in deep tech
World Food Program (WFP) – UN agency; Hunger Map Live, smart warehousing, supply chain optimization
Bayer Crop Science – Agrochemical; AI-driven FPO surveillance, precision farming
Lead Connect – AI analytics startup; soil, crop, climate modeling
BL Agro Industries – Agriculture business; AI systems integration
Axis Bank – First bank globally certified ISO 42001
FIKI – India agriculture/innovation body (multiple task forces)
Bharat Innovation Fund – Deep tech venture fund
MIT Naneda – Multi-agent internet architecture (referenced for Kumbh Mela AI pilots)

Technical Concepts & Resources

AI/ML Standards & Frameworks

ISO 42001 – AI Management System standard (analogous to ISO 27001 for security)
ISO 4219 – AI standards family
NIST AI Risk Management Framework (AI RMF v2) – Safe, secure, resilient, valid, appropriate, verifiable
ML Commons / AI Illuminate – Global benchmark for practical AI standards
ISO 4271 – Security management (referenced for analogy)

Assurance & Governance Methodologies

Responsible AI Framework – Microsoft's six principles: fair, reliable & safe, private & secure, inclusive, accountable, transparent
Risk-based governance – Tiered assurance based on use case consequence/materiality
Living inventory – Continuous catalog of AI systems in use
Continuous monitoring & retirement – Post-deployment oversight

Agricultural AI Technologies & Concepts

Precision farming – Plot-by-plot, data-driven interventions (fertilizer, pesticide, irrigation)
Direct Seeded Rice (DSR) – Low-water, low-methane cultivation method
Farmer Producer Organizations (FPOs) – Aggregation structures for collective decision-making
Federated learning – Model training across distributed data without centralized data sharing
Remote sensing / satellite imagery – Hyperspectral data, microwave, SAR, thermal imaging for crop monitoring
Yield prediction – 10m × 10m granular forecasting before harvest
Soil sensors & weather stations – IoT data collection (2 lakh planned under government initiatives)
Drones for surveillance – Real-time crop health monitoring
Carbon trading systems – Incentivizing low-emission cultivation
Chatbots & advisory systems – Multi-language, farmer-accessible guidance (Bhart Vistar)

Agricultural Data Infrastructure & Initiatives

Bhart Vistar – Government AI chatbot for farmer advisory (multi-language)
Agree Stack – Public data infrastructure for agriculture
Agree Course – Agricultural data aggregation (Ministry of Agriculture)
PMISAN – PM Support Scheme; ~10 crore farmers monthly data entry
ABDM (Ayushman Bharat Digital Mission) – Health data sovereignty; referenced as model for farmer data control
Unified Farmer ID (UFI) – Biometric farmer identity linked to land records, credit, subsidies (8 cr+ farmers registered; targeting 100% in 1 year)
Digital Locker – Trusted system for document storage (India DPI)
Krishi Niti (Agricultural Policy) – Government guidance on tech deployment

AI Architecture Concepts

Agentic AI – Autonomous/semi-autonomous agents that plan, use tools, communicate with other agents
A2A (Agent-to-Agent) – Communication protocols
MCP (Model Context Protocol) – Simpler communication standard
Large Language Models (LLMs) – OpenAI's GPT, Anthropic's Claude, OpenAI's models
Small Language Models (SLMs) – Context-aware, task-specific, lower compute
Edge computing – Local CPU/device execution vs. cloud-dependent
Digital Sandboxes – Shared testing environments for cross-border AI validation
Strong AI architectures – Multi-domain integration (soil, water, atmosphere, landscape) for ecosystem-level problems
Bayesian architecture – Probabilistic, interpretable (vs. black-box approaches)

Key Data Sources & Monitoring Systems

Hunger Map Live (WFP) – Real-time satellite-based food insecurity tracking
PMKSY (Pradhan Mantri Krishi Sinchayee Yojana) – Irrigation data
IMD (India Meteorological Department) – Weather forecasting
MONET – Monk Museum AI project (painting discovery system, not generative)
Kumbh Mela pilot – Multi-agent internet pilot for pilgrim experience economy (MIT Naneda)
Monastic painting collection – 90% in vault; AI system enabling discovery of 3% in gallery

Regulatory & Governance References

EU AI Act – European regulation framework
UK AI Governance Framework – Principles-based approach
Indian AI Governance Guidelines – Published during summit; establishes AI Governance Group and Tech/Policy Expert Committee
Singapore AI governance – Regional framework for assurance integration
GDPR – Data privacy (referenced in context)

Public Digital Infrastructure (DPI) Concepts

UPI (Unified Payments Interface) – India's successful DPI model; API-based, cost-reducing, ecosystem-enabling
Aadhaar – Biometric identity infrastructure (reference for Unified Farmer ID scaling)
API economy – Open, modular interface-based ecosystem growth
Digital twin – Virtual replicas of physical systems (mentioned: digital twins of ocean, supply chains)

Organizational Capabilities & Use Cases Mentioned

Organization	Focus	Key Capability
CINTF	Cross-border AI collaboration	Shared testing environments, interoperability standards, federated research
TCS	Knowledge work democratization	Bridging elite/average skill gaps; cultural institution partnerships (Monk museum)
Nokia	Modular open systems	Ecosystem orchestration; open interfaces; standardization (telco + AI)
Applied Ventures	Deep tech supply chains	Semiconductor, venture investing, multi-country fund co-development
Microsoft	Responsible AI operationalization	M365 Copilot; ISO 42001 certification; internal governance at scale
Holistic AI	Enterprise AI governance	SaaS platform; agentic AI testing; continuous monitoring; small/medium business support
BSI	Standards & certification	ISO 42001 auditing; medical device review; first certification providers
WFP	Food security through AI	Hunger prediction; supply chain optimization; warehouse robotics; safety net strengthening
Lead Connect + BL Agro	Precision agriculture	Soil/crop/climate modeling; multi-district deployment (155 districts); value chain transformation
Bayer Crop Science	FPO-scale mechanization	Drone surveillance networks; plot-level advisory; export-grade traceability
Arya	Agricultural credit & storage	Post-harvest value chain; market linkage; farmer aggregation
Neteb Software	Agricultural data platforms	Small model deployment; edge computing; DPI integration; cost-reduction