Asia Needs a Common Framework to Measure Trust in AI, New Report Warns

Asia Society Policy Institute calls for shared metrics as fragmented national strategies risk deepening regional inequalities and stalling adoption

A sweeping new analysis of artificial intelligence governance across 15 Asian countries has found that while nearly every nation in the region invokes “trust” as central to its AI strategy, none has a reliable or consistent way to measure it, a gap that experts warn could undermine the continent’s ambitions as it races to harness AI for economic growth.

The report, published by the Asia Society Policy Institute (ASPI), proposes nine measurable factors, or metrics, for what it terms “trusted AI ecosystems,” covering everything from data quality and compute infrastructure to misinformation governance and environmental sustainability. Its findings draw on policy analysis and two roundtable discussions held on the sidelines of the 2026 AI Impact Summit in New Delhi, involving experts from Australia, Hong Kong, India, Indonesia, Japan, Malaysia, Singapore, South Korea, Sri Lanka, the United Arab Emirates, and the United States.

A Race with High Stakes

The economic stakes behind the region’s AI push are considerable. UNESCO’s AI readiness assessments project suggests that widespread adoption could add up to USD 1.9 trillion to India’s GDP by 2035 and USD 113.4 billion to Malaysia’s economy by 2030. Indonesia’s national strategy frames AI adoption as the path to developed-country status by 2045.

Yet the report finds that this urgency is running ahead of governance. Regulatory frameworks across Asia remain calibrated largely around model-level risks, even as AI development has moved toward agentic systems, autonomous agents that access, combine, and act on data across jurisdictions. The report’s authors warn that governance frameworks unable to measure trust in static deployments are “even less equipped to track it across dynamic, multi-actor agentic architectures.”

Nine Factors, One Framework

ASPI’s proposed framework breaks trust down into nine domains:

1. Trusted datasets. Many Asian nations acknowledge the importance of open government data, but few have addressed whether that data is actually ready for AI training, properly labeled, structured, interoperable, and representative of diverse populations. The report identifies a three-layer trust problem: trust between government ministries (a challenge Indonesia candidly calls “ego-sectoral” attitudes), trust between the state and citizens regarding data collection and consent, and the technical coherence of the data itself. India’s AI Kosh initiative and Singapore’s “data concierge” mechanism are cited as early models, though the report notes that most countries hold ambitions that “far outpace tangible progress.”
2. AI infrastructure. Nations across Asia are building data centers and expanding cloud services, but geopolitical tensions, particularly between the United States and China, have intensified concern about dependence on foreign providers. Smaller economies like Bhutan and Nepal are positioning their renewable energy resources as a potential competitive edge in hosting energy-efficient infrastructure, while countries such as Indonesia and Sri Lanka continue to face foundational shortfalls in domestic computing capacity.
3. AI skills and awareness. The report draws a sharp distinction between the challenge of developing frontier technical talent, where South Korea and Taiwan are investing in semiconductor expertise and AI chip programs, and the broader challenge of preparing low-skill workforces in populous economies like Bangladesh and the Philippines for automation-driven disruption. Across the region, reskilling ambitions are common, but concrete curricula and measurable targets are rare. The roundtables recommended creating an Asia AI Knowledge Facility to pool process knowledge and enable peer learning.
4. Global AI value chain leverage. Most of Asia remains downstream in the global AI supply chain, dependent on hardware and compute it does not control. China’s dominance in processing rare earth minerals gives it structural leverage over every other nation’s AI ambitions. Indonesia’s experience as the world’s leading nickel producer offers a cautionary note: despite pursuing downstream processing, Chinese firms still control roughly 75% of refining capacity. The report argues that trust in Asia’s AI ecosystem depends partly on “managed interdependence, not pure self-sufficiency.”
5. Ethical AI development. Across the region, ethics guidelines are common; enforceable ethics law is rare. Japan’s flagship AI legislation takes a soft-law, voluntary compliance approach. South Korea stands out with its AI Framework Act, which includes penalties for breaches and an AI ethics committee, though it has drawn criticism from start-ups who say it favors foreign firms. Singapore has built a layered system of sector-specific guidelines across finance, healthcare, and generative AI. Several countries, including Indonesia and Brunei, ground their ethical frameworks in national philosophies or religious principles rather than imported international standards.
6. Misinformation governance. China has enacted the region’s most operationalized response with a 2025 law mandating both implicit and explicit labeling of AI-generated content. South Korea’s Framework Act includes fines for non-compliance with content labeling requirements. The Philippines adopted sector-specific rules during its 2025 elections, requiring disclosure of AI-generated campaign materials. The report notes, however, that some governments may have incentives to exploit AI-enabled misinformation rather than combat it, framing the issue as “a public health problem with a range of responses.”
7. AI governance frameworks and institutions. The region broadly favors “pro-innovation, pro-safety” governance, principle-based, risk-tiered, and sector-sensitive, over sweeping punitive regulation. But institutional capacity varies enormously. Malaysia’s governance has been fragmented across multiple ministries; the country recently created a National AI Office to coordinate efforts. The report calls for “iterative governance” with clear roles for developers, deployers, users, and regulators, and urges that AI governance be reframed around the full lifecycle of agentic AI systems, including post-deployment monitoring.
8. Environmental sustainability. AI’s environmental footprint is growing rapidly. International Energy Agency data cited in the report estimates that data center electricity consumption represented approximately 1.5% of global electricity use in 2024, a figure projected to reach 4.4% by 2035. The report finds that most Asian national strategies treat this challenge as peripheral. Bhutan’s 2025 National AI Strategy is highlighted as an exception for explicitly naming environmental concerns. A coordinated systems approach proposed at the AI Impact Summit calls for energy proportionality, infrastructure assessed against environmental criteria, carbon transparency frameworks, and standardized cross-border metrics.
9. Cybersecurity. Threats range from AI-driven financial crime and phishing (a priority for Singapore, Thailand, and Bangladesh) to vulnerabilities in critical information infrastructure in less-developed digital economies. The report warns that governance discourse has been disproportionately focused on generative AI misuse, synthetic content, and deepfakes, while neglecting model security, data integrity, supply chain vulnerabilities, and systemic resilience. One roundtable speaker described an attempt by a commercial entity to replicate a large language model through systematic prompt injection, illustrating how adversarial AI behavior now extends well beyond nation-states and criminal groups.

A Call for Regional Coordination

A central argument running through the report is that Asian countries are largely pursuing “sovereign” AI strategies in isolation, competing for investors rather than pooling resources, and often replicating, rather than reducing, their dependencies on external players.

The report does not argue that national differences should be erased. Rather, it calls for a common analytical foundation: baseline interoperable principles on safety, accountability, and transparency; mutual recognition pathways for audits and incident response; cross-border data-sharing arrangements; and regular regulator-to-regulator cooperation.

Roundtable participants also recommended that rather than constructing new trust indices, policymakers should extract trust-relevant components from existing governance, digital readiness, and cybersecurity indices that are already being collected, but not yet being used to systematically assess trust.

“Trust cannot be generated in isolation,” the report concludes, “particularly in economies deeply embedded in global markets and technology dependencies.” Without shared metrics, cross-country comparisons remain abstract, and effective practices continue to be shared on an ad hoc basis, leaving one of the world’s most consequential technology transitions without a common map.

Other People are Reading