Principles for ethical artificial intelligence are easy to display and harder to put into practice. UNESCO‘s fourth Global Forum on the Ethics of AI, held in Riyadh, placed that gap at the centre of discussion. In a summary published on 18 September 2026, the organization reported more than 6,300 participants and delegations from over 50 Member States. Attendance does not measure successful governance, but it shows that the issue has moved beyond a narrowly technical debate. The more consequential news is elsewhere: UNESCO presented several instruments intended to help public administrations identify gaps and turn shared values into policy decisions.
A 2021 recommendation meets the test of implementation
The forum came five years after the 2021 adoption of UNESCO’s Recommendation on the Ethics of Artificial Intelligence. The text sets out human dignity, fundamental rights, transparency, fairness, human oversight and environmental sustainability among its guiding concerns. An international recommendation, however, is not a national law that enforces itself. To shape a system used in a school, hospital or public office, it needs capable institutions, dependable data, oversight mechanisms and people able to demand accountability. The Riyadh meeting framed this phase of implementation as its priority.
The event was co-hosted by UNESCO and Saudi Arabia, represented by the Saudi Data & AI Authority and the International Centre for AI Research and Ethics. It was the first edition of the global forum in the Gulf region, following meetings in Prague, Kranj and Bangkok. Geography matters: AI policy is not made in a single centre, and countries have very different capacities. Yet an international forum remains a place for exchange, not a regulator. Its statements and tools will need to be judged by adoption, use and results, not just by who appears in a group photograph.
RAM 2.0: diagnosis before promises
One of the main launches announced was RAM 2.0, a more action-oriented version of UNESCO’s AI Readiness Assessment Methodology. Introduced in 2023, the RAM helps countries examine their ability to develop and govern AI responsibly, looking at institutions, rules, skills and other conditions for implementation. Its purpose is to reveal gaps, not to issue certificates of good behaviour. The forum also presented a meta-analysis bringing together 55 country reports. A synthesis on that scale can reveal recurring patterns; it cannot automatically make very different political and economic systems directly comparable.
A useful diagnosis must lead to choices. Where should an independent authority be strengthened? Which public bodies can afford to audit a tool? How can someone challenge an automated decision that affects them? These are possible extensions of the ethical framework, not new measures UNESCO imposed in Riyadh. Keeping the distinction clear guards public debate against two shortcuts: assuming a document alone can govern a market, or dismissing assessment because it imposes no penalty. A method’s value will be visible in the budgets, procedures and lines of responsibility it helps clarify.
Putting the environment into the calculation
Another resource unveiled in UNESCO’s official summary is a toolkit addressing AI’s effects on the environment and ecosystems. Forum discussions considered the energy, water and material resources required by AI systems. Claims of efficiency cannot therefore be assessed without their physical costs. A service may improve one task while shifting consumption to less visible data centres or supply chains. The toolkit is intended to help policy-makers account for such consequences. The announcement does not say that it sets an emissions ceiling or already provides a universal measurement for every model’s footprint.
That caution matters for a technology spanning small specialized tools and systems requiring vast infrastructure. Comparing their impacts calls for explicit boundaries and verifiable data. UNESCO also announced action cohorts and expert training intended to adapt tools to concrete projects. Training people, again, is a beginning, not proof that every administration now has sufficient resources. Scaling up will depend on staff continuity, independence and the ability to publish assessment results, including uncomfortable ones.
Cooperation tested where AI is used
UNESCO’s report also describes a ministerial segment involving 52 delegations. Discussions covered education, health, culture, accessibility, languages and AI safety, alongside the environment. This range shows why AI ethics cannot be reduced to an IT compliance checkbox. Language policy, a patient’s access to an explanation and teachers’ involvement in choosing a tool are social as well as technical decisions. The challenge is to keep those affected in view when the conference ends and procurement, legislation and evaluation begin.
For B-EMPIRE, the Riyadh story comes down to a practical question: who can show that the announced tools really change practice? The 2021 recommendation offers a common vocabulary; RAM 2.0, the synthesis of 55 reports and the environmental toolkit seek to give it administrative form. Their presentation and the forum’s scale are confirmed facts. Their effectiveness, funding and ability to protect people remain open. The goal is not to declare AI ‘ethical’ at the end of a summit. It is to make decisions explainable, contestable and revisable wherever the technology is deployed.
