UNESCO Launches First Global Comparison of AI Readiness Across 55 Nations

Day 2 of the 4th UNESCO Global Forum on the Ethics of AI opened with a landmark: the first-ever comparable dataset on how countries are governing artificial intelligence, drawn from 55 national readiness assessments. The session revealed a striking paradox — 82% of countries are integrating AI into education, but only 13% are training educators on AI ethics — and heard from ministers in small and developing nations who are using the tool to chart their own paths.

• UNESCO releases meta-analysis of 55 countries' AI readiness, launches RAM 2.0 for standardized, evidence-based governance assessment
• Ministers from Comoros, Somalia, Suriname, and Uganda share first results from piloting the updated readiness tool
• Global panel concludes AI ethics must be embedded across entire government organizations, not confined to tech teams
• Africa's CAFRAD director calls for a binding African Charter on AI Ethics and sovereign AI systems built with African critical minerals
• Saudi Arabia details end-to-end national AI infrastructure: data bank, 480 MW data center, and one-million-citizen training initiative

---

From Principles to Data: RAM 2.0 Arrives

Why it matters: For the first time, AI governance data from 55 countries — predominantly from the Global South — can be compared side by side, giving policymakers a baseline to measure progress rather than relying on anecdotes or reports from wealthier nations.

Where things stand: Dafna Feinholz, Chief of Bioethics and Ethics of Science and Technology Section, UNESCO, opened the session by framing the Readiness Assessment Methodology as the bridge between the 2021 UNESCO Recommendation on the Ethics of AI and on-the-ground implementation. "We have already 78 countries, which is 40% of UNESCO's membership, implementing the RAM. And that's really important," she said.

The meta-analysis, led by Julia Pomares, Co-Founder and Regional Lead, LATAM Chapter, The Global Initiative for Digital Empowerment, synthesizes findings from those national reports. The topline numbers are sobering: while a strong majority of countries are moving AI into classrooms, the infrastructure for teaching people how to think critically about AI barely exists. Only 30% of countries report policies addressing the gender gap in the AI workforce — a figure that rises to 43% only in formal recommendations, not in practice.

Feinholz underscored why the data matters beyond the usual headlines. "Normally we have information about what happens in AI in the world in reports that are coming from countries of the Global North, but we don't have much information about what's happening on the global majority," she said. "Thanks to this meta-analysis that includes a lot of countries from the global majority, we have information firsthand of that situation."

Marc-Antoine Dilhac, Director of Governance and International Collaboration, OBVIA, who led the development of RAM 2.0, characterized the upgrade as a fundamental shift in methodology. "We need less narratives, more evidence-based assessments. And we are shifting from standalone reports with RAM 1.0 to interoperable data with RAM 2.0," he said. RAM 2.0 introduces stronger alignment with the Recommendation's policy areas, conditional pathways for easier implementation, and new coverage of environmental impact, gender, Indigenous peoples, and persons with disabilities.

Dilhac also pushed back on the idea that a single global template can simply be dropped into any country. "The translation that matters is not linguistic. The translation that matters is how do we translate a universal norm, which is the recommendation, into a local institutional and cultural grammar," he said. Countries currently piloting RAM 2.0 include Uganda, Somalia, Comoros, Suriname, Zambia, Sri Lanka, and Quebec.

---

Five Countries, Five Realities: The RAM on the Ground

Why it matters: The ministerial panel demonstrated that the readiness tool is not an academic exercise — ministers are using it to write national AI strategies, prioritize spending, and make political arguments about who should benefit from AI.

Uganda: From Digital Infrastructure to AI-Specific Ethics

Joyce Nakatumba Nabende, Senior Lecturer, Makerere University and Member, UN Independent International Scientific Panel on AI, explained that RAM 2.0's gate-question structure allowed Uganda to distinguish its AI-specific assessment from prior digital infrastructure audits. The tool helped the country focus on cultural diversity, gender equality, Indigenous groups, and environmental impact — dimensions that existing frameworks had not surfaced. The findings are feeding directly into Uganda's forthcoming National AI and Emerging Technology Strategy.

Comoros: 900,000 People, 23 Recommendations, and 270 Girls

Minister Mmadi Hassani Oumouri, Minister of Posts, Telecommunications, the Digital Economy, and Transparency, Comoros, delivered one of the session's most pointed interventions. Comoros produced 23 recommendations organized in three implementation waves through 2030, prioritizing a national AI strategy, multi-stakeholder governance, and data protection. But Minister Oumouri challenged the assumption that small nations need the same infrastructure as large ones: Comoros, with roughly 900,000 inhabitants, does not require a Tier 3 data center on the Saudi scale — it needs to invest in its people.

He highlighted a program training 270 girls aged 6–18 in AI and cybersecurity, and made a broader ethical argument: "J'ai la forte conviction qu'on peut arriver vite à l'éthique de l'intelligence artificielle si cette technologie, cette innovation est dans les mains des femmes et filles," he said — expressing his strong conviction that AI ethics is more attainable when the technology is in the hands of women and girls.

Somalia: Local Languages, Local Models

State Minister Ahmed Osman Dirie, State Minister of Communications and Technology, Somalia, described how RAM maps the AI ecosystem across five dimensions, helping Somalia align infrastructure growth with ethical governance. He stressed the importance of developing local language AI models, noting that "the current technological advancements are driven by the West and they are based on foreign languages and foreign cultures." Building human capital and ensuring digital inclusion were central themes.

Suriname: An Ongoing Process, Not a Snapshot

Minister Andrew Bassarone, Minister of Economic Affairs, Entrepreneurship and Technological Innovation, Suriname — the first Dutch- and English-speaking Caribbean country to complete RAM 2.0 — convened more than 100 participants from over 60 institutions. He framed the tool as iterative rather than static: "I think the process of assessing, acting, measuring, and again reassessing will make it possible for us to keep this process going and making sure that RAM is not just a picture that we take, a momentary picture, but that it's a process ongoing, evolving." Suriname faces the distinctive challenge of balancing AI-driven economic growth with the preservation of 90% rainforest coverage.

Saudi Arabia: Closing Gaps Identified in Round One

Abdurrahman Habib, Senior Strategic Advisor, Saudi Data and AI Authority (SDAIA), confirmed Saudi Arabia's intent to take RAM 2.0 to update its assessment and continue closing gaps identified in the first round.

---

The Training Gap: Ethics Can't Be a Side Module

Why it matters: Research presented at the panel found that government employees with less AI training are actually more confident in AI outputs — meaning deploying AI tools without embedding ethics training creates a governance risk, not just a knowledge gap.

Where things stand: Moderated by Idil Kaner, Senior AI Policy Fellow, Center for AI and Digital Policy (CAIDP), the institutional capacity panel assembled perspectives from the Middle East, Africa, Saudi Arabia, the United Kingdom, and Jamaica.

Akmaral Orazaly, Principal Researcher, Mohammed Bin Rashid School of Government, presented survey data from 327 AI companies across 10 MENA countries showing early-stage ethical implementation concentrated in soft mechanisms — ethics officers and internal policies — with limited third-party audits. The more alarming finding was behavioral: "People with less training and less competence in AI systems have more confidence in AI outputs. So if we only focus on training AI fluency without actually helping employees understand judgment and how to critically assess the outputs of AI systems, we run the risk of having employees or companies that can work with AI but are not safe."

Natalia Domagala, Founder, Uncovering Algorithms, drew on UK experience to distill five lessons: AI capacity must include ethics skills; ethics must be embedded across all teams, not just technical ones; principles need practical mechanisms like ethical impact assessments; feedback loops must exist; and organizations need a culture of critical challenge with engaged senior leadership. "We can no longer talk about AI skills without AI ethics skills," she said.

The other side: Coffie Dieudonné Assouvi, Director General, CAFRAD, outlined Africa's position in three pillars: recognizing AI's development potential, developing sovereign African AI systems, and believing in international cooperation. He called for a binding African Charter on AI Ethics — "une norme contraignante" — modeled on what the EU and Council of Europe have established. His most direct challenge: "Le premier défi de l'Afrique aujourd'hui, c'est d'avoir une IA africaine, made in Africa, avec les minerais critiques de l'Afrique" — Africa's first challenge is to have AI that is made in Africa, with African critical minerals. He argued that AI not built for Africa's linguistic, cultural, and developmental context poses heightened risks, and that soft skills — critical thinking, judgment, discernment — matter as much as technical competencies.

Mamdouh Alenezi, General Manager, SDAIA Academy, described Saudi Arabia's three-level capability model: training individuals (one million Saudis trained, 55% women), tailoring programs per agency, and ensuring data privacy through national data management offices.

Nadine Matthews Blair, CEO, Crescent Advisory and Advisory Board Chair, Jamaica Artificial Intelligence Association, highlighted the particular challenges facing small island developing states, where technology advances faster than regulatory capacity. She pressed for practical flexibility: "What does right-sizing in terms of recommendations, or what could that look like so that there are more fit-for-purpose recommendations that allow smaller nations and larger nations and less-resourced nations and highly-resourced nations to still implement something and have meaningful impact?"

---

Saudi Arabia's National AI Blueprint

Hotham Altuwaijri, Deputy Director, National Center for AI (NCAI), SDAIA, delivered a keynote detailing Saudi Arabia's institutional AI infrastructure. Key elements include the National Data Bank for governing and integrating government data; a data maturity index measuring governance quality across agencies; the Hexagon data center in Riyadh with 480 MW capacity; a national AI adoption index (NAIB) tracking agency-level implementation; and the SAMAI initiative training one million citizens in AI.

Altuwaijri emphasized that institutional readiness is a long-term strategic investment, and that success should be measured by tangible impact on institutional efficiency, decision quality, and responsiveness to citizens — not the number of AI models deployed. The presentation also covered national AI ethics principles and a risk management framework aligned with data privacy regulation.

---

• UNESCO-Oxford partnership announced offering free, globally accessible AI governance training for public servants worldwide.