APAC Centre for AI: Regional Leadership in a Global AI Economy

Summit Talk Summary

---

Key Topics Covered

• Regional AI Opportunities & Ecosystem Building

  • Applied AI at sectoral levels (BFSI, healthcare, government, public services)
  • Startup ecosystem challenges (capital access, VC wait-and-watch mentality)
  • Corporate-startup-academic collaboration models

• Cross-Border Partnerships & Market Access

  • Soft landing programs and reciprocal international pathways
  • Trade, tariff, and export policy improvements
  • Interoperable sandboxes and regulatory coordination

• Data Governance & Policy Harmonization

  • Data quality, representativeness, and bias in datasets
  • Cross-border data sharing constraints
  • Fragmented regulatory frameworks across APAC jurisdictions

• Evaluation & Benchmarking

  • Eurocentrism in current AI safety benchmarks
  • Need for culturally contextual evaluation metrics
  • Low-resource language and region-specific performance testing

• Infrastructure & Shared Enablers

  • Open-source datasets and multilingual data resources
  • Interoperable regulatory sandboxes
  • Digital public goods (e.g., Wikipedia, Wikidata models)

• Trust & Security in Cross-Border AI

  • Federated learning approaches
  • Interoperable trust frameworks for payments and transactions
  • Cybersecurity awareness across tier-2/tier-3 cities

• Financing & Investment Models

  • Deep tech funding gaps and capital constraints
  • Government research funding initiatives (ANRF, RDF, AI Mission)
  • Sovereign LLMs and open-source alternatives to closed Western systems

• Policy & Governance Frameworks

  • Risk-based evidence for high-risk AI use cases
  • Decentralized-centralized governance (60% common + local variations)
  • Language dataset quality and linguistic diversity

• Accessibility & Inclusivity

  • Digital literacy and AI literacy programs
  • Reaching farmers, villages, and underserved communities
  • Democratization of AI tools beyond coding expertise

• Responsible AI & Ethical Scaling

  • Inclusivity and community-centered data collection
  • Addressing existing societal data gaps (e.g., women in healthcare)
  • 360-degree approach to ethical frameworks

---

Key Points & Insights

1. Applied AI Over Technology-First Approaches
   Dr. Shibu (Aquarians AI) emphasizes that startups and corporates are shifting from pure technology narratives (e.g., "agentic frameworks," vectorized databases) to problem-solving and business impact focused on specific sectors (BFSI, healthcare, government services). The center should facilitate this use-case-driven mentality.

2. Decentralized-Centralized Governance Model
   Rather than imposing uniform global frameworks, the APAC center should develop 60% common governance standards across the region with local variations. Agreement on what constitutes "high-risk AI use cases" is the foundational prerequisite.

3. Deep Tech + Domain Expertise = Success
   Startups combining deep technical innovation with domain-specific knowledge are most likely to succeed. Capital access remains the primary constraint for seed-stage deep tech companies.

4. Data Representativeness & Bias as Entry Points for Systemic Bias
   Gita (Karya AI) identifies that current datasets lack linguistic and cultural diversity across APAC languages and contexts. Since bias embeds at the data-collection stage, building representative datasets is critical governance work.

5. Eurocentrism in AI Benchmarks & Evaluation
   Current safety benchmarks and evaluation metrics are rooted in Western/Eurocentric contexts and do not account for APAC socioeconomic realities, low-resource settings, or region-specific biases. New evaluation frameworks must test cultural relevance and regional context.

6. Public-Private-Academic Partnerships as Structural Enablers
   Vun (Mastercard India) and Jeremy (Ahmedabad University) converge on the need for structured partnerships that provide startups with access to technology, models, and best practices, while corporates gain implementation partners and governments gain strategic direction. Awareness-building is an underestimated component.

7. Translational Labs Bridge Academic Research to Market
   Dr. Shibu notes the emergence of translational research labs positioned between academic research (T1–T3 scale) and production systems (T7–T9). Universities are beginning to incentivize patents, licensing, and commercialization alongside publications.

8. Trust Deficit in Global Institutions Requires Regional Solutions
   Mayor Ramachandra emphasizes that 50 years of global institutional trust structures are collapsing. Responsible AI and trust frameworks are not optional—they are foundational to building new regimes. India's emphasis on responsible AI is strategically crucial.

9. Open-Source & Sovereign Models Over Closed Western Systems
   Rather than competing with proprietary US/Chinese systems, India and APAC should invest in sovereign LLMs, open-source models, voice-led technologies, and multilingual AI that address regional problems (e.g., digital public goods, agricultural innovation).

10. Inclusive, Community-Centered Data Governance
    Sonia Basker emphasizes that responsible AI frameworks must actively address inclusivity gaps (e.g., women in healthcare datasets) and prevent the digitalization of bias. Digital AI literacy and accessibility to underserved communities (farmers, villages, non-English speakers) are structural prerequisites.

---

Notable Quotes or Statements

• Sha Tari (Plug and Play India): "The way we do things today will just be much more efficient. We don't have to worry about losing jobs; we have to in fact be happy that we will be much more efficient."

• Dr. Shibu (Aquarians AI): "No one wants to talk to you about technology anymore... the prominent question across all these areas [is] What is the problem you're trying to solve? What business impact are you going to create?"

• Jeremy (Ahmedabad University): "India's given a lot more than it's received from its overseas partners. We need to ensure that the frameworks for soft landings are reciprocal."

• Mayur Ramachandra (Salmander Advisers): "There is a huge deficit in trust [in global institutions]. The last 50 years whatever we have put together in terms of systems and structures it's being dismantled. Somebody has to step in and put in a new regime, and that is where responsible AI and trust are absolutely crucial."

• Gita Kandar (Karya AI): "Embedding datasets is the first point of entry where bias starts embedding in the system. It's very important that we ensure these policies are part of the governance framework."

• Vun Sakhuja (Mastercard India): "[We can build] crossborder interoperable trust frameworks because trust is the bedrock of anything in payments. The reason we operate in 200+ countries is because there is trust that if a transaction happens, the merchant will get paid."

• Pravin Das (Wikimedia Foundation): "Walking towards a multilingual internet is something we should aim at, providing solutions to billions of users not speaking English."

• Aditi Sitha (Dutient): "One of the most valuable roles which the APAC Centre can play is that of a policy harmonizer because there's a lot of fragmented AI frameworks running currently."

• Tripati Bansil (Policy Index AI): "India and APAC are very diversified. We need to be cognizant from the beginning that we need to play into these diversities and not make it uniform."

---

Speakers & Organizations Mentioned

Name	Role/Title	Organization
Sha Tari	Program Lead	Plug and Play India
Dr. Shibu (Shivarama Krishnan)	Chief AI Scientist	Aquarians AI
Jeremy	CEO of Venture Studio	Ahmedabad University
Mohit Jen	Principal Researcher	Microsoft Research India
Gita Kandar	Head of Policy & Government Engagement	Karya AI
Vun Sakhuja	Government Engagement Director	Mastercard India
Mayur Ramachandra	Founding Partner	Salmander Advisers
Pravin Das	Lead Partnerships Manager	Wikimedia Foundation
Kamesh (Moderator)	Co-organizer	Core AI Coalition
Sonia Basker	Public Policy Consultant & Trainer	(Independent)
Tripati Bansil	Founder & CEO	Policy Index AI
Aditi Sitha	Principal Consultant	Dutient.ai
Rama Shipkumar	Founder (SRK Game Changers)	Starting Innovation Labs
Vidit	Senior Engineering Manager	Adobe (launching startup)
Amit Mitt	CEO	Aviation Company (GIFT City-based)

Institutions & Initiatives Referenced:

• Core AI Coalition (multistakeholder coalition on AI in India)
• GIB City (GIFT International Financial Services Centre)
• Ahmedabad University (Department of Science & Technology)
• Startup India (government initiative)
• DPIIT (Department for Promotion of Industry & Internal Trade)
• NASSCOM (National Association of Software & Services Companies)
• DARPA, NSF (US research funding)
• ANRF, RDF (Indian research schemes)
• AI Mission (Indian government initiative)
• Quantum Mission (Indian government initiative)
• ICIO (International Civil Aviation Organization, UN)

---

Technical Concepts & Resources

• Agentic AI / AI Agents: Framework for autonomous AI systems; mentioned in context of workflow automation in BFSI, healthcare, and government services.

• Knowledge Management & Digital Twins: Integration with agentic frameworks for sector-specific applications.

• Vectorized Databases: Infrastructure for embedding-based search and retrieval in AI applications.

• Federated Learning: Privacy-preserving distributed machine learning model; referenced as approach to cross-border AI collaboration and interoperable trust frameworks.

• Translational Labs (T1→T3→T7→T9 pipeline): Academic research → translation to market. T1–T3 = fundamental/translation research; T7–T9 = production systems.

• Sovereign LLMs (SLMs): Language models built and owned by nations/regions to reduce dependence on US/Chinese proprietary systems.

• Interoperable Sandboxes: Regulatory testing environments (e.g., GIFT City's model aggregating data from RBI, SEBI, IFSCA) allowing startups to build solutions acceptable across multiple jurisdictions.

• Soft Landing Programs: Structured pathways for startups to pilot and commercialize in foreign markets with reciprocal access to funding, best practices, and regulatory clarity.

• Multimodal AI / Voice-Led AI: Technologies prioritizing speech and voice interfaces for accessibility in low-literacy and non-English speaking populations.

• Digital Public Goods: Open-source infrastructure (Wikipedia, Wikidata, datasets) available for reuse across regions.

• Risk-Based Evidence / Risk Assessment Frameworks: Standardized methodologies for evaluating high-risk AI use cases and determining jurisdiction-specific compliance.

• Benchmark & Evaluation Frameworks:

  • Current benchmarks noted as Eurocentric
  • Needed: culturally contextual metrics, low-resource language performance, region-specific bias testing

• MONTHAN (Government of India): Platform indexing private and public pilot opportunities for researchers; facilitates pathways to commercialization.

• AI4C (AI Code): India AI's open-source AI co-creation initiative.

• Knowledge Systems (Indian Knowledge Systems / IKS): Integration of traditional Indian knowledge with AI for agriculture, crafts, and cultural heritage.

• Humans in the Loop (Netflix film): Documentary highlighting cultural and contextual gaps in current AI datasets.

---

Policy & Governance Frameworks Discussed

• Data Sovereignty & Cross-Border Data Sharing: Identified as major constraint; current regulations prevent easy sharing of sensitive health, financial, legal data across borders.

• Decentralized-Centralized Governance: 60% common regional standards + local variations (vs. global uniformity).

• High-Risk AI Use Case Categorization: Region must agree on what constitutes high-risk before building harmonized frameworks.

• Regulatory Sandboxes: Interoperable testing environments (e.g., GIFT City model) allowing simultaneous compliance across multiple jurisdictions.

• Cybersecurity Awareness Standards: Gap identified in tier-2/tier-3 cities; awareness workshops reveal even basics (password hygiene, OTP protection) are underemphasized.

• Trade & Tariff Harmonization: Identified barriers to cross-border startup scaling; ongoing discussions with EU, UK, Australia.

• Export Policies & Talent Mobility: ESOPs and startup talent require clearer pathways for regional movement.

---

Identified Gaps & Roadblocks

1. Capital Access: Deep tech seed-stage startups face "wait and watch" VC mentality; grants and equity funding remain underutilized.
2. Policy Fragmentation: Data localization, explainability, and compliance requirements differ across APAC jurisdictions.
3. Data Representativeness: Datasets lack APAC linguistic diversity, cultural context, and underrepresented demographics (e.g., women in healthcare).
4. Evaluation Eurocentrism: Current benchmarks not rooted in socioeconomic realities or cultural relevance of APAC regions.
5. Academic-to-Market Translation: Researchers incentivized for publishing, not commercialization; IP and patents not flowing to startups/enterprises.
6. Trust Deficit: Collapse of global institutional trust; new regional frameworks needed.
7. Jurisdictional Risk: Unclear liability, compliance, and regulatory responsibility across borders (especially for sensitive sectors like aviation, healthcare).
8. Accessibility Gap: Rural, low-literacy, non-English-speaking populations underserved by current AI solutions.
9. Collaboration Fragmentation: Indian institutes of prominence (IITs, IIM, research bodies) siloed; "criminal wastage" of overlapping efforts.
10. Visibility/Awareness: Innovations in APAC not sufficiently visible to corporates, governments, or international audiences.

---

End of Summary