194 States, 1,200 Delegates, One Question: UNESCO's AI Ethics Forum Opens in Riyadh

Today, September 14, 2026, the doors of the Ritz-Carlton in Riyadh open on the fourth UNESCO Global Forum on the Ethics of Artificial Intelligence. Over four days — through September 17 — some 1,200 policymakers, regulators, and researchers representing all 194 UNESCO member states will work through more than 30 sessions and 100+ speakers under a theme that reads like a diagnosis: “Transforming Global Cooperation for Ethical AI Governance.”

The forum is convened by UNESCO, hosted by Saudi Arabia’s data and AI regulator SDAIA, and co-organized with ICAIRE, the International Center for Artificial Intelligence Research and Ethics headquartered in Riyadh. On paper, it is a policy conference. In practice, it is the largest single checkpoint in the slow, uneven project of turning the world’s most widely adopted AI ethics text into something with teeth.

Why this forum matters more than the last three

UNESCO’s Recommendation on the Ethics of AI was adopted by acclamation in November 2021 by all 193 member states then seated — still the only global, normative AI framework with that kind of reach. The problem has never been the principles. Proportionality, fairness, transparency, sustainability, human oversight: nearly every government signs on to those words. The problem is the gap between a principle and a verdict — between “AI systems should be fair” and a regulator, often with a staff of a dozen, determining that a deployed benefits-eligibility model is not.

That gap is precisely what this edition of the forum is structured around. The 2026 program pairs a Scientific Briefing and closed-door ministerial meeting with seven high-level dialogues, 21 expert panels, and two side-event tracks run by ICAIRE and SDAIA. UNESCO is also using the opening to launch new practical tools for ethical AI governance — building on the Ethical Impact Assessment (EIA) instrument it has been refining, which walks governments and developers through assessing the ethical implications of an individual AI system before and during deployment.

The GNAIS story: regulators building shared muscle

Perhaps the most consequential thread in Riyadh this week runs through GNAIS — the Global Network of AI Supervisory Authorities. Launched at the 2025 forum in Bangkok, GNAIS grew out of an informal working group of European AI regulators that UNESCO scaled into a global network. Its chair? Saudi Arabia’s SDAIA — the same body hosting the forum.

GNAIS exists because its members share one awkward inheritance: they are now responsible for supervising systems that, until recently, were nobody’s job to check. Their mandates, legal powers, and technical capacity vary wildly. What they can standardize is methodology — a defined way of turning “fairness” into a measurable check that produces the same verdict whether it is run in Brussels, Nairobi, or Riyadh.

A workshop led at the forum by Dr. Gemma Galdon-Clavell, founder of audit firm Eticas, pushes exactly that agenda: what independent AI evaluation looks like in practice, and what supervisory authorities can standardize across borders so they aren’t each reinventing supervision from scratch. Two of its working assumptions deserve attention. First, evaluation has to happen where the system actually runs — a model that passes a pre-deployment test can still drift once it meets real applicants, patients, or claimants. Second, a one-time check is not oversight: systems change, populations change, and the risk profile a regulator signed off on a year ago may no longer hold. Continuous monitoring against established benchmarks is the emerging consensus.

What the agenda says about the state of play

Read the Day 1 plenary agenda as a diplomatic document and the priorities are unmistakable. The morning moves from opening remarks by SDAIA President Dr. Abdullah bin Sharaf Alghamdi and UNESCO Deputy Director-General Åsa Regnér through a fireside chat on “AI, Multilateralism and Diplomacy” featuring the co-chair of the UN’s Global Dialogue on AI Governance — a direct line to the UN General Assembly process that has been debating AI governance resolutions since 2024.

The afternoon belongs to two anniversary conversations: “Five Years of the Recommendation: Where We Are, Where We Go Next,” spread across three panels stacked with ministers from Cameroon, Argentina, Ghana, Djibouti, Montenegro, Paraguay, and Tanzania, among others; and a two-part plenary on “Financing and Partnerships for Ethical AI Governance for the People,” where the speakers list runs from Iran’s science minister to Sberbank’s human-centric AI lead to the CEO of the Saudi National Infrastructure Fund. The National AI Index (NAII) awarding ceremony — an effort to benchmark national AI readiness — sits between them.

That financing track is not decoration. One recurring critique of global AI governance talk is that it produces principles for rich countries and platitudes for everyone else. UNESCO has pushed Readiness Assessment Methodology (RAM) exercises across dozens of member states, and the forum’s emphasis on capacity building, regional cooperation panels spanning the Gulf, Africa, and Latin America, and funding mechanisms is an acknowledgment that a governance framework without financing is a PDF.

The venue question

No honest coverage of this forum can skip the irony in its location. Saudi Arabia chairs GNAIS, hosts ICAIRE, and is spending tens of billions — through HUMAIN and its infrastructure programs — to become a top-tier AI power, while its record on digital expression and surveillance draws sustained criticism from human-rights organizations. UNESCO’s defense is instrumental: the states shaping AI’s next decade include governments the multilateral system needs inside the tent, and the Recommendation’s own standards — including strong provisions on data protection and human rights — apply to hosts as much as anyone.

That tension is the forum’s real stress test. Global cooperation on AI ethics only means something if the network being built in rooms like the Ritz-Carlton’s can evaluate systems neutrally — including those deployed by the governments paying for the room.

What to watch

Three outcomes would mark this edition as more than a four-day communique. First, adoption of the new governance tools: if the EIA and Observatory-linked instruments get taken up by national authorities as standard practice rather than shelf-ware, the principle-to-verdict gap narrows measurably. Second, GNAIS deliverables on cross-border evaluation methodology — shared testing standards are the difference between 50 regulators each guessing and a genuine supervisory network. Third, money: concrete financing commitments for ethical AI capacity building in the Global South would answer the loudest critique of the entire enterprise.

The last three forums — in Prague, Ljubljana, and Bangkok — built the vocabulary. Riyadh’s job is to build the machinery.