Riyadh Puts AI Governance on the Global Stage

The fourth UNESCO Global Forum on the Ethics of Artificial Intelligence took place in Riyadh from September 14 to 17, 2026. UNESCO and the Kingdom of Saudi Arabia co-hosted the forum, represented by the Saudi Data and AI Authority and the International Center for AI Research and Ethics. Its official theme was “Transforming Global Cooperation for Ethical AI Governance.”

The forum brought together participants from government, civil society, science, academia, international organisations and the private sector. Its stated purpose was to connect ethical principles with practical governance and implementation.

From principles to implementation

The programme covered AI readiness and institutional capacity, supervision and accountability, public administration, inclusion, education, health, culture, sustainability, energy, safety and international cooperation. These were the forum's stated areas of discussion, not evidence that every governance challenge has already been solved.

UNESCO's 2021 Recommendation on the Ethics of Artificial Intelligence provided the policy foundation. According to UNESCO, more than 75 Member States have engaged with its Readiness Assessment Methodology to support implementation of that recommendation.

Engagement with a methodology is a meaningful step and a modest one. It indicates that a government has examined its own institutional capacity. It does not indicate what that government then changed.

Governance is becoming part of deployment

Our reading is that AI governance is moving closer to the operational layer. Institutions increasingly need processes for assessment, oversight, accountability and ongoing monitoring, not only high-level principles.

That distinction matters. A forum can establish shared language and exchange practical approaches, but responsible deployment still depends on what governments and organisations implement after the event.

It also changes who does the work. Principles are written by policy teams. Monitoring, logging, escalation paths and incident review are built by engineering and risk functions, and they carry a running cost that someone has to fund.

A forum can establish shared language and exchange practical approaches, but responsible deployment still depends on what governments and organisations implement after the event.

The distance between a recommendation and an obligation

UNESCO's instrument is a recommendation. It sets a shared standard and invites Member States to align with it. It does not, by itself, create enforceable duties on a company deploying a system in a given market.

That gap is where most of the practical questions sit. Which existing regulator supervises an automated decision in health, credit or employment? Does liability attach to the organisation that built the model or the one that deployed it? What is a person's route to redress when an automated decision goes against them, and does it work without a lawyer?

These are not criticisms of the forum, which was convened to advance exactly this conversation. They are the questions by which its results will eventually be judged, and they are answered in national law rather than in plenary sessions.

Why Riyadh matters

Riyadh's role as host placed the Gulf inside an international discussion that has often been associated with Washington, Brussels, London and major technology centres. The location does not by itself prove leadership in AI governance. It does show that Saudi institutions are investing in convening international policy and technical stakeholders around the subject.

Convening power is real power. It shapes which problems are framed as urgent and whose definitions are used. Whether it converts into regulatory influence depends on what the host subsequently adopts at home.

The signal

The strongest signal from Riyadh is the shift in emphasis from agreeing that responsible AI matters to examining how it can be governed in practice.

For companies, governments and investors, trust will depend on evidence: how systems are assessed, who is accountable, what safeguards apply and whether those safeguards continue to work after deployment.

The next edition of this forum will be easier to assess than this one, because by then there will be a record to check the language against.