Governments are shifting from discussing AI ethics in theory to implementing concrete risk assessment practices for public services. The International Telecommunication Union (ITU) Academy is launching a specialized training program that equips policymakers and practitioners with structured methodologies to identify, evaluate, and mitigate risks in AI-enabled government systems.

Why Are Governments Struggling to Translate AI Ethics Into Practice?

As AI and data-driven systems reshape public service delivery across justice, health, education, and social protection, governments face a critical gap: they understand that fairness, transparency, and accountability matter, but they lack practical tools to assess these risks before deploying systems that affect millions of citizens. The ITU course addresses this directly by moving beyond general principles and awareness to applied impact assessment methodology.

The program, supported by the European Union's Global Gateway strategy, is designed for government officials, policymakers, and development practitioners who directly design, procure, regulate, or oversee AI-enabled systems. It's particularly relevant for agencies working in sensitive domains where algorithmic decisions carry high stakes for vulnerable populations.

What Framework Are Governments Using to Assess AI Risk?

The course builds on the United Nations Development Programme's (UNDP) Human Rights Impact Assessment (HRIA) methodology, adapted specifically for AI and digital governance. Rather than treating AI ethics as a compliance checkbox, this framework guides participants through a structured process to scope a use case, identify risks and impacts, and develop mitigation and governance recommendations.

Participants work in teams throughout the five-session course to develop a draft impact assessment and mitigation plan for a real-world digital or AI governance use case relevant to their institutional contexts. By the final session, they present actionable recommendations grounded in their own government environments.

Steps to Conduct a Practical AI Impact Assessment

• Scope the System: Define the specific AI or data-driven system being deployed, including its intended use, affected populations, and decision-making scope within government services.
• Identify Affected Stakeholders and Impacts: Map who is affected by the system's decisions and what potential harms or benefits could result, including fairness risks, transparency gaps, and exclusion of marginalized groups.
• Conduct Risk Analysis: Evaluate ethical risks such as bias and autonomy violations, as well as human rights risks including discrimination, lack of accountability, and denial of due process.
• Develop Mitigation Measures: Design safeguards, accountability mechanisms, and policy responses that address identified risks before or during system deployment.
• Establish Governance Recommendations: Create institutional oversight structures, audit processes, and escalation procedures to ensure ongoing compliance and responsiveness to emerging harms.

What Specific Risks Are Governments Learning to Identify?

The course curriculum explicitly addresses two categories of risk that often go undetected in government AI deployments. Ethical risks include fairness, bias, autonomy violations, and transparency failures. Human rights risks encompass discrimination, exclusion, lack of accountability, and violations of due process.

These distinctions matter because a system might appear technically fair on paper while still violating human rights principles in practice. For example, an AI system used in immigration decisions might show statistical parity across demographic groups but still exclude asylum seekers from meaningful review of algorithmic determinations. The HRIA methodology helps governments catch these gaps.

Who Is Eligible, and What Will Participants Learn?

The program is limited to 35 participants and targets government officials with at least two years of experience in digital policy, ICT, or governance. Applicants must hold an undergraduate degree or have three years of relevant professional experience, demonstrate fluent English, and submit a motivation letter explaining how they plan to apply the tools in their institutional context.

Upon completion, participants will be able to explain the role of impact assessment in responsible digital transformation, identify key risks in AI and data-driven systems, understand ethical and human rights assessment approaches, apply a structured methodology to real-world use cases, and propose practical mitigation and governance measures.

The course runs across five two-hour sessions beginning in October 2026, with a one-week break between the fourth and final sessions. Participants are graded on pre-class quizzes, attendance and participation, and a group project that produces a draft impact assessment and mitigation plan. A score of 70 points or higher earns an ITU certificate.

Why Does This Matter for Citizens and Governments?

When governments deploy AI systems without rigorous impact assessment, citizens can face algorithmic decisions that are biased, unexplainable, or unaccountable. A hiring algorithm used by a labor ministry might systematically disadvantage women. A health system's resource allocation AI might deny care to underserved regions. A criminal justice risk assessment tool might perpetuate historical discrimination.

By equipping policymakers with practical assessment tools, the ITU course aims to prevent these harms before systems go live. The methodology ensures that governments ask hard questions about who benefits, who is harmed, and what safeguards are needed before deploying AI in public services that affect fundamental rights.