Asia Society Policy Institute

• Establish and prioritize local datasets, and incentivise government organisations, social and private enterprises to share datasets and collaborate on the structure of those sets. The next wave of innovation will be in small model development. The governments should take lead in designing clear incentive structures and governance mechanisms. Such incentives will ensure that local actors see value in participation, while AI companies will perceive commercial viability. • Governments should play a critical role in fostering a conversation about the value of user data and its protection. Trust will suffer if end users either feel they are exposed or exploited. Governments should develop standards for data privacy and the United Nations can help foster the global discussion on rights and protections of personal data, including when it can be licensed or sold and how those profits are distributed. Doing so will ensure that control of personal data becomes part of our constellation of human rights, ensuring dignity, transparency and inclusion. • Establish an Asia AI knowledge facility that provides stakeholders AI process knowledge, including understanding data orchestration, algorithmic development, sectoral workflows, regulatory bottlenecks, and institutional constraints. This enables the development of context-relevant AI solutions that can be adopted responsibly and at scale.

• Interoperability of governance approaches
• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• AI capacity-building

Trust is critical to AI adoption, and adoption is a prerequisite for economic transformation anchored to the digital technologies. To establish trusted ecosystems across Asia, trust has to be designed into systems (through safeguards, explainability, grievance redress, and accountability), established in practice (via transparent processes, measurable metrics, and audits), and maintained over time (through continuous oversight as models and contexts evolve). Within the broader technology ecosystem, immediate challenges to address include enabling interoperable datasets, incentivising data sharing and empowering governments and end users on AI process knowledge. Trusted data sets are foundational in realizing the transformative potential of AI tools, and governments hold key data sets that are not yet available to AI developers. Three challenges exist with regard to unlocking this data: 1. There are few incentives inside governments for its own departments or ministries to share information with each other. This must become a high level political priority with a structure of interagency cooperation to overcome bureaucratic politics. 2. There is an absence of structured data sets for specific use cases and end users. Involving end users directly in the process of model development will help explain the potential and the limitations of current AI models, and create trust among those with little exposure to AI tools. Inclusivity in model development also aligns incentives. 3. For AI to solve the existing global challenges and be adopted at scale, stakeholders should have the requisite AI process knowledge, including understanding sectoral workflows, regulatory bottlenecks, and institutional constraints.

ASPI's research proposes nine factors that together define the conditions that national strategies must get right to enable responsible and rapid AI adoption: trusted datasets, AI infrastructure, AI skills and awareness, global AI value chain leverage, ethical AI development, misinformation governance, AI regulation and governance institutions, environmental sustainability, and cybersecurity. Each factor represents a domain where the absence of measurable progress creates conditions that can stall adoption, erode public confidence, or concentrate power in ways that undermine the broader ecosystem.

Consensus building and creation of a common metrics, terminology on AI governance is the most important aspect of the global dialogue. There should be a platform to build trusted reliance network.

1. Participants concurred that governments should consider creating an Asia AI knowledge facility that includes information about data orchestration, algorithmic development, sectoral workflows, regulatory bottlenecks, and institutional constraints. 2. There is a need to foster structured partnerships that are anchored on incentive-based data sharing among AI private enterprises, civil society organizations, and government institutions. 3. The impact of AI development and adoption is diverse in terms of globally spread talent, discreet in terms of algorithm development, and distributive in terms of hardware and cloud services. To responsibly develop and ensure equal access to AI across Asia, countries must have a common framework to measure both safety of and trust in an AI ecosystem. 4. consider establishing an AI oversight board that acts as an independent adjudication board, ensuring that the AI companies adhere to the global standards, protect the protect the rights of individuals, respect intellectual property, and preserves human autonomy. 5. Governments should establish iterative governance rather than rigid legislation with defined roles and relationships between actors (developers, deployers, users, regulators), establishing AI sandboxes to test the outputs against the identified risks, and constant stakeholder consultations.

Governments should establish iterative governance rather than rigid legislation with defined roles and relationships between actors (developers, deployers, users, regulators), establishing AI sandboxes to test the outputs against the identified risks, and constant stakeholder consultations. The priority in the iterative AI governance should be to create doctrinal clarity in intellectual property, data governance, and liability law, particularly in contexts involving autonomous or semi-autonomous agentic behaviour. This will shape incentives, accountability mechanisms, and public trust over the long term, aiding the predictable market in Asia. Importantly, there is a need to reframe AI regulation to align with the layered architecture of AI and enable the broader technology ecosystem. It should include energy generation, digital connectivity, access to AI infrastructure, robust back-haul networks, data acquisition and model training to deployment, orchestration by AI Agents, and post-deployment monitoring. Complementing the latter, a sustained, cross-government institutional capacity should be built to govern the systemic risks and strategic opportunities these systems generate. AI governance should be reshaped to include "agentic AI stack," where AI agents' access, combine and act on personal, enterprise and operational data. It's time to reframe AI governance around the full lifecycle and layered architecture of agentic systems, and invest in robust, cross-government capacity to manage these emerging risks and opportunities across the value chain.