Governments worldwide are deploying artificial intelligence to improve public services, but the officials responsible for overseeing these systems often lack the knowledge to do so effectively. A critical governance blind spot is emerging, not within the technology itself, but within the institutions tasked with managing it. UNESCO's AI Literacy Training for Civil Servants reveals that public officials frequently arrive expecting to learn how to use AI tools, only to discover they need to understand ethics, risk management, and accountability instead.

What Is the "Institutional Black Box" in AI Governance?

While policymakers focus heavily on the technological "black box" problem, a parallel challenge is being overlooked: the institutional black box. When civil servants lack sufficient understanding of AI systems, they cannot ask informed questions, recognize risks, or make responsible decisions about procurement, deployment, and oversight. The human oversight chain in government extends far beyond a single reviewer approving an AI recommendation. It includes the policymaker who approves an AI initiative, the procurement officer evaluating vendor proposals, the regulator assessing compliance, and the manager determining whether an AI system is appropriate for public service delivery.

Without adequate AI literacy among these officials, the concept of "human-in-the-loop" becomes merely procedural rather than meaningful. Yomna Omran, International and Government Affairs Manager at the Personal Data Protection Center in Egypt and a trainer for UNESCO's program, noted that participants often finish the training with more questions than they started with, reflecting a growing awareness of the complexity of AI governance.

"Rather than feeling that they have completed their learning journey, they often leave with a greater appreciation of how much there is still to learn," Omran explained.

Yomna Omran, International and Government Affairs Manager, Personal Data Protection Center – Egypt

How Are Governments Building AI Governance Capacity?

Recognizing this gap, several governments are taking concrete steps to strengthen AI governance capabilities among public officials and institutions. Thailand, Singapore, and Hong Kong are leading efforts to move from voluntary frameworks toward comprehensive regulatory approaches, while simultaneously investing in workforce development (Sources 1, 2, 3).

• Thailand's Comprehensive Approach: Thailand released a revised draft AI Act on July 2, 2026, adopting a risk-based framework influenced by international standards like the EU AI Act. The government also announced expanded AI governance training programs, scaling from 40 initial trainers to 3,000 by 2027, and launched an AI Governance Practice Centre to serve as a regional implementation hub.
• Singapore's Assurance Ecosystem: Singapore is launching an AI Tester Accreditation Programme in Q3 2026, the first of its kind in Asia, to accredit third-party companies that test and evaluate AI systems. This builds on the AI Verify framework launched in 2022, which helps companies assess responsible AI implementation against 11 internationally recognized governance principles.
• Hong Kong's Supervisory Intensification: Hong Kong's Privacy Commissioner launched compliance checks on 60 organizations in January 2026, with results published in May 2026 showing that 95% used AI in daily operations. The regulator recommended governance structures, privacy impact assessments, AI audits, staff training, and incident-response plans.

These initiatives reflect a broader global pattern. AI regulation remains fragmented across jurisdictions, but common themes are emerging: risk-based controls, transparency, human accountability, incident management, safety testing, data governance, cybersecurity, and assurance.

Why Does Localization Matter in AI Governance Training?

UNESCO's experience delivering AI literacy training revealed that the approach is most effective when delivered in the language and context where public officials actually work. AI governance discussions are often conducted in English and rely on case studies from a limited number of contexts, which can reduce their practical impact. When adapted to local realities, the same training becomes significantly more relevant to officials' daily responsibilities.

This localization principle extends beyond language. Officials in Egypt, Tunisia, and other regions benefit from examples grounded in their own governance structures and policy environments. What initially appears as theoretical discussion about ethics and accountability gradually becomes practical conversation about decisions already being made across ministries and public institutions.

What Specific Governance Challenges Are Emerging in Asia-Pacific?

As AI deployment accelerates, regulators in Singapore and Hong Kong are identifying new risk categories that require specialized oversight. Agentic AI, a category of systems capable of autonomous, multi-step decision-making with limited real-time human involvement, has emerged as a distinct governance challenge.

Singapore's Infocomm Media Development Authority launched the world's first Model AI Governance Framework for Agentic AI in January 2026, updated in May 2026 with real-world case studies and best practices from over 50 organizations. The framework emphasizes that humans remain ultimately accountable for AI agent outputs and recommends both technical controls, such as least-privilege access and logging mechanisms, and non-technical measures, including clear accountability structures and meaningful human oversight.

Hong Kong's Privacy Commissioner issued a specific alert on agentic AI in March 2026, identifying it as an elevated privacy risk. AI agents may access local devices, files, emails, credentials, and browser contents, and may autonomously execute tasks without real-time user involvement. The regulator recommended restricting AI systems to minimum access rights, avoiding administrator privileges, and maintaining human-in-the-loop for decisions with significant individual impact.

Financial sector oversight has also intensified. Singapore's Monetary Authority published the MindForge AI Risk Management Operationalisation Handbook in March 2026, providing detailed guidance for financial institutions adopting AI. For agentic AI specifically, the handbook identifies risks including unauthorized actions, cascading errors across connected systems, data breaches, and governance scalability challenges.

What Are the Key Takeaways for Organizations and Governments?

The convergence of these developments signals a fundamental shift in how governments approach AI governance. The focus is moving from developing ethical principles and voluntary frameworks toward operationalizing governance that works across sectors and institutions. Thailand's AI Governance Week 2026 emphasized this transition from "policy to practice," with the government announcing mandatory AI governance requirements across public agencies and enforcement mechanisms to ensure compliance.

For organizations developing or deploying AI systems, the regulatory landscape is hardening. Thailand's draft AI Act, if enacted, would introduce obligations extending beyond Thailand's borders, establish compliance requirements for higher-risk AI systems, impose transparency obligations for AI-generated content, and create one of the most comprehensive AI regulatory regimes in the Asia-Pacific region. Foreign AI providers serving individuals in Thailand may be required to appoint a local representative with broad authority to act on their behalf.

The broader lesson is clear: effective AI governance depends not only on technical safeguards and regulatory frameworks, but on building the institutional capacity and human expertise needed to implement them. Without civil servants, regulators, and organizational leaders who understand AI governance, even the most sophisticated regulatory frameworks will fail to achieve their intended purpose.