UNITED NATIONS | HIGH-LEVEL ADVISORY BODY ON AI

Governing AI for Humanity

Final Report of the UN Secretary-General’s High-Level Advisory Body on Artificial Intelligence

Published: 2024

Mandate: 38 experts | 68 countries

Key finding: 118 countries excluded from 7 major AI initiatives

Executive Summary

The UN Secretary-General’s High-level Advisory Body on AI, comprising 38 multidisciplinary experts from governments, civil society, private sector, and academia, was convened to analyze and advance recommendations for international AI governance. After 18 deep dives, over 250 written submissions, and consultations across all regions, the final report Governing AI for Humanity identifies a critical global governance deficit: a fragmented patchwork of norms and institutions where 118 countries are excluded from major AI governance initiatives. The report proposes seven actionable recommendations structured around common understanding, common ground, and common benefits, plus a UN AI office as a lightweight coordination mechanism. This case study provides a comprehensive analysis of the report’s findings, recommendations, expert risk perceptions, and implications for global AI policy up to 2030, with a special focus on Monitoring & Evaluation professionals.

The Global AI Governance Deficit

The report finds that despite hundreds of AI governance documents, frameworks, and principles adopted by governments, companies, and international organizations, there is a fundamental disconnect between high-level rhetoric and the conditions required for safety, equity, and inclusiveness. Accountability is often absent, compliance rests on voluntarism, and the governance landscape remains nascent and full of gaps.

118

Countries party to zero of seven major non-UN AI governance initiatives

Countries party to all seven major initiatives (mostly Global North)

74%

Experts expect pace of AI technical change to accelerate (Risk Pulse Check)

71%

Experts concerned that AI harms will become substantially more serious

Three Critical Governance Gaps

Representation Gap

Entire regions excluded from AI governance conversations. Analysis of 7 prominent initiatives shows severe geographic imbalance: Africa, Latin America, and large parts of Asia absent.

Coordination Gap

Fragmented initiatives risk incompatible AI governance regimes. Even within UN system, 27+ instruments touch on AI but none address AI holistically.

Implementation Gap

Commitments fail to translate into tangible outcomes. Missing enablers: capacity, funding, accountability mechanisms.

The Seven UN Recommendations for Governing AI for Humanity

The report advances a holistic vision organized around three pillars: common understanding, common ground, and common benefits — supported by coherent effort.

Pillar A: Common Understanding

Recommendation 1: International Scientific Panel on AI

Establish an independent, multidisciplinary scientific panel under UN auspices (modelled partly on IPCC). Mandate: annual reports on AI capabilities/opportunities/risks; quarterly thematic digests on AI and SDGs; ad hoc reports on emerging risks.

Pillar B: Common Ground

Recommendation 2: Policy Dialogue on AI Governance

Launch twice-yearly intergovernmental and multi-stakeholder policy dialogue on AI governance at UN. Purpose: share best practices, promote interoperable governance, share AI incident information.

Recommendation 3: AI Standards Exchange

Create clearing house for AI standards bringing together ISO, IEC, ITU, IEEE, tech companies, civil society. Tasks: register of definitions and standards, identify gaps.

Pillar C: Common Benefits

Recommendation 4: AI Capacity Development Network

Network of collaborating UN-affiliated centres providing expertise, compute, AI training data. Sandboxes, online education, fellowship programmes.

Recommendation 5: Global Fund for AI

Independent governance structure receiving financial/in-kind contributions. Provides shared computing resources, sandboxes, benchmarking tools, SDG model repository.

Recommendation 6: Global AI Data Framework

Establishes definitions, principles, common standards for AI training data provenance, data trusts, model agreements, and cross-border data interoperability.

Pillar D: Coherent Effort

Recommendation 7: UN AI Office (Secretariat)

Light, agile AI office within UN Secretariat reporting to Secretary-General. Acts as glue supporting all other proposals, engages stakeholders, advises Secretary-General.

Key AI Risks Identified by the UN Body (Risk Global Pulse Check)

Information integrity: 78% experts concerned

Wealth/power concentration: 74% concerned

Armed conflict AI use: 75% concerned

Discrimination: 67% concerned marginalized communities

Autonomous weapons: 120 Member States support treaty

Environmental costs: energy consumption of AI systems

Implications for Monitoring & Evaluation (M&E) Professionals

New Evaluation Domains

AI fairness and bias audits
Algorithmic accountability frameworks
Human rights impact assessments of AI systems

Skills for Evaluators

AI technical literacy
Data governance evaluation
Digital and AI readiness assessments

Governance Indicators

AI policy effectiveness metrics
Capacity development tracking
Multi-stakeholder governance evaluation

Key Lessons from the UN AI Governance Process

Lesson 1: Technology without governance increases inequality. Innovation alone does not guarantee inclusion; without deliberate governance mechanisms, AI benefits concentrate in already-advantaged countries.

Lesson 2: Global public goods need global institutions. Just as climate change required the IPCC, AI requires coordinated governance — lessons from IAEA, ICAO, CERN, and FATF are instructive.

Lesson 3: Capacity building is essential. Countries cannot regulate or benefit from AI without expertise, infrastructure, and data.

Lesson 4: Multi-stakeholder governance works best. Governments alone cannot govern AI effectively — collaboration among governments, academia, civil society, and private sector is required.

Analysis and Critique by EvalCommunity

Disclaimer: This case study presents an independent analysis of the United Nations High-Level Advisory Body on Artificial Intelligence final report Governing AI for Humanity, led by the UN Secretary-General and with contributions from 38 global experts. It is based on publicly available information from the UN (official report, consultations, and annexes including Risk Global Pulse Check and Opportunity Scan). This analysis reflects the perspectives of EvalCommunity on strengths, gaps, and recommendations for M&E practice in AI governance, technology-facilitated policy evaluation, and global AI accountability frameworks. EvalCommunity does not claim ownership of the original UN content. We encourage readers to consult the original UN report for complete context and verbatim recommendations.

Strengths identified: The report’s framing around representation, coordination, and implementation gaps provides a robust diagnostic tool for evaluators. The proposal for an International Scientific Panel on AI (Recommendation 1) offers a clear precedent from IPCC that M&E professionals can adapt for impact assessment. The Global Fund and Capacity Development Network (Recommendations 4 and 5) address critical missing enablers that evaluators have long identified as barriers to equitable technology adoption.

Gaps and recommendations for M&E: While the report mentions accountability and monitoring, it lacks concrete metrics for tracking progress on the seven recommendations. EvalCommunity recommends developing a standardized AI Governance M&E Framework with indicators for representation (e.g., percentage of Global South countries in policy dialogues), coordination (e.g., number of interoperable standards adopted), and implementation (e.g., funds disbursed to low-income countries). Additionally, evaluators should advocate for baseline assessments of AI readiness in excluded countries before capacity-building interventions begin.

Reflections on a Future International AI Agency

The report does not currently recommend a full international AI agency with enforcement powers, but acknowledges that if risks become more acute, such an institution may become necessary. Thresholds could include: development of uncontrollable AI systems, deployment of systems untraceable to human actors, or emergence of superintelligence capabilities. Functions could draw from IAEA, OPCW, ICAO, CERN, and FATF: monitoring, verification, compliance, emergency response, and promoting peaceful uses.

Frequently Asked Questions

What is the High-Level Advisory Body on AI?

The High-Level Advisory Body on Artificial Intelligence was convened by the UN Secretary-General in 2024 to analyze and advance recommendations for international AI governance. It consisted of 38 diverse experts from governments, civil society, private sector, and academia, balanced by geography, gender, and discipline.

What is a global AI governance deficit?

A global AI governance deficit means governance mechanisms are nascent, fragmented, and full of gaps. Accountability is often absent, compliance rests on voluntarism, and 118 countries are excluded from major AI initiatives despite being affected by AI.

References & Further Reading

United Nations. (2024). Governing AI for Humanity: Final Report of the High-Level Advisory Body on Artificial Intelligence. New York: United Nations.
UN General Assembly Resolution 78/265 (2024). Seizing the opportunities of safe, secure and trustworthy AI systems for sustainable development.
UN High-Level Advisory Body on AI. (2023). Interim Report: Governing AI for Humanity.
Official report download: Governing AI for Humanity (PDF)
EvalCommunity Academy: Courses on AI in Monitoring and Evaluation

Explore more resources on AI governance and M&E

Visit EvalCommunity Academy →