Table of Contents
When public officials register for an artificial intelligence (AI) training program, some arrive with a particular expectation. They anticipate learning how to write better prompts, automate routine tasks, or use the latest generative AI tools to increase productivity. Some are surprised—and occasionally disappointed—to discover that the training focuses instead on AI ethics, governance, risk management, and responsible use.
Yet an interesting transformation often takes place during the course of the program.
As discussions move beyond the technology itself and into questions of accountability, human oversight, public trust, transparency, bias, procurement, and decision-making, participants begin to connect these issues to their own professional responsibilities and to the decisions they make every day as public officials. What initially appeared to be a theoretical discussion about ethics gradually becomes a practical conversation about governance.
By the end of UNESCO’s AI Literacy Training for Civil Servants, many participants offer a similar reflection: the program covered far more than they expected, introduced issues they had never previously considered, and revealed the complexity of questions surrounding 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.
This experience highlights an important but often overlooked dimension of responsible AI governance. While significant global attention has been devoted to developing ethical principles, regulatory frameworks, and technical safeguards, less attention has been paid to the people responsible for implementing, overseeing, procuring, regulating, and ultimately governing AI systems.
If AI systems are frequently described as “black boxes,” one of the most significant governance challenges may not lie within the technology itself, but within the institutional blind spots that emerge when those responsible for oversight lack the knowledge and confidence required to exercise it effectively.
Looking Beyond the Technical Black Box
The “black box” problem has become one of the defining metaphors in discussions about AI. Policymakers, researchers, and practitioners rightly seek greater transparency and explainability in increasingly complex systems whose outputs may influence important decisions.
Yet focusing exclusively on the technological black box risks obscuring another challenge: the institutional black box.
Around the world, governments are exploring how AI can improve public services, increase administrative efficiency, support policy design, and enhance decision-making. Civil servants are increasingly asked to evaluate AI solutions, participate in procurement processes, oversee implementation, assess risks, and develop governance frameworks. In some cases, they may also be responsible for monitoring the impacts of systems already in use.
This reality raises an important question: who are the humans expected to remain “in the loop”?
The answer extends far beyond the individual reviewing an AI-generated recommendation. In the public sector, the human oversight chain begins with the decision to procure, deploy, regulate, or monitor an AI system in the first place.
The policymaker who approves an AI initiative, the procurement officer evaluating vendor proposals, the regulator assessing compliance, the manager determining whether an AI system is appropriate for a particular public service, and the official responsible for monitoring outcomes are all parts of that chain.
Human oversight, therefore, is not simply a technical safeguard embedded within an AI system. It is a capacity challenge for the institutions that govern it. Meaningful
oversight depends on people possessing sufficient understanding to ask informed questions, challenge assumptions, recognize risks, and make responsible decisions. Without that capacity, the concept of “human-in-the-loop” risks becoming procedural rather than meaningful.
What AI Literacy Training Revealed
The experience of delivering UNESCO’s AI Literacy Training for Civil Servants revealed several lessons about how public officials engage with AI governance.
One of the most striking observations was the gap between initial expectations and eventual takeaways. Many participants arrived expecting practical guidance on AI tools and productivity applications. For some, a program focused on ethics and governance appeared less immediately relevant than learning how to use the latest technologies. This perception rarely lasted long.
As discussions became grounded in real-world public sector scenarios, participants began to recognize that responsible AI was not a separate topic sitting alongside AI adoption—it was inseparable from it. Questions about bias became relevant to public service delivery. Discussions about transparency became relevant to accountability. Human oversight became relevant to decisions already being made across ministries and public institutions. What initially seemed theoretical quickly became practical.
Another recurring observation was that participants often finished the program with more questions than they had at the beginning. Topics such as AI governance, risk management, procurement, accountability, and oversight frequently generated discussions that participants felt deserved deeper exploration in dedicated training programs of their own.
Rather than indicating a lack of understanding, this reflected a growing awareness of the complexity of AI governance and the range of issues public institutions must navigate. The objective of AI literacy is not to turn every civil servant into an AI specialist. It is to equip public officials with the knowledge and confidence needed to engage critically with AI-related decisions and responsibilities.
One of the clearest indications of this shift came during a discussion of Egypt’s AI governance ecosystem. I remarked that the next time they read a news article about a meeting of the National Council for Artificial Intelligence, they would likely
see it through a different lens. Rather than viewing it as another government announcement, they would better understand the governance structures, coordination mechanisms, and policy decisions behind it. What was once a headline had become a governance process they could now recognize and interpret.
Localization Matters: Speaking the Language of Governance
The experience also demonstrated that AI literacy is most impactful when delivered in the language and context in which public officials perform their daily work.
AI governance discussions are often conducted in English and frequently draw upon examples, case studies, and policy debates originating from a limited number of contexts. While these resources provide valuable insights, their impact can be significantly enhanced when adapted to local realities.
Through UNESCO’s AI Ethics Experts Without Borders (AIEB) network, the training program was designed to reach countries in their own languages and contexts. This approach involved far more than translating training materials. It required localization.
Concepts had to be explained using terminology familiar to public officials. Examples needed to resonate with national priorities and institutional realities. Discussions had to reflect local governance structures, public sector challenges, and policy environments.
The Arabic version of the training provides one example of this broader approach. In the cohorts I delivered in Egypt and Tunisia, participants were able to engage with AI governance concepts in the language they use in their professional responsibilities, while also exploring examples tailored to their national and institutional contexts. This significantly enhanced engagement and helped bridge the gap between global principles and local implementation.
The experience reinforced an important lesson: responsible AI governance is not about importing external frameworks unchanged. It is about adapting shared principles to diverse national realities. In this sense, localization itself becomes a governance enabler.
From AI Literacy to Institutional Readiness
AI literacy alone will not solve every governance challenge associated with artificial intelligence. However, it provides a foundation upon which broader institutional readiness can be built. It is not an end state, rather, it lays the groundwork for more advanced governance capabilities by enabling public officials to engage with the tools, methodologies, and institutional processes that help translate responsible AI principles into operational practice.
As governments move from experimentation to implementation, they will increasingly require governance mechanisms capable of supporting responsible AI adoption. These may include risk assessment methodologies, procurement guidance, internal governance structures, cross-government coordination mechanisms, monitoring processes, and communities of practice that facilitate continuous learning.
However, governance mechanisms alone cannot ensure responsible AI adoption. Their effectiveness ultimately depends on the people responsible for implementing them. Risk assessment frameworks, procurement guidance, oversight structures, and monitoring processes are only as effective as the institutional capacity supporting them.
The experience of AI literacy programs suggests that public officials are eager to engage with these issues when given the opportunity. Far from viewing responsible AI as an obstacle to innovation, many participants recognized it as an essential component of trustworthy and effective AI adoption.
As governments seek to balance innovation with public trust, strengthening public sector capacity remains a core component of responsible AI governance. The experience of UNESCO’s AI Literacy Training for Civil Servants demonstrates how international cooperation, local expertise, and contextualized delivery can help translate global principles into practical governance capabilities.
READ MORE: https://www.unesco.org/ethics-ai/en/articles/beyond-black-box-reducing-governance-blind-spots-through-ai-literacy-civil-servants?hub=701
