Building practical skills for AI supervision
At the 4th UNESCO Global Forum on the Ethics of AI, held in Riyadh, Saudi Arabia, from 14 to 17 September 2026, GNAIS held a workshop on AI audit and evaluation led by Dr Gemma Galdón-Clavel, Usman Gohar, and Alex Maagard from Eticas.ai. The workshop combined presentations on AI auditing with a practical exercise to give participants hands-on experience of the audit process.
The session introduced the Eticas AI Risk Taxonomy and explored how broad AI risks can be translated into concrete elements that can be assessed. The discussion focused on the importance of moving from identifying a risk to defining how that risk can be tested, what evidence should be collected and how the findings can be communicated.
Participants were introduced to a structured approach that breaks risks down into more specific mechanisms that can be tested. The workshop also discussed the limitations of simple pass or fail assessments, which can make it difficult to understand the severity of a finding or where a system requires further attention.
Participants also went through a practical case involving an AI-based hiring tool. This case illustrated different aspects of an AI system as part of an audit, including the data used by the system, how the model is coded, how outputs are produced and checked, and how responsibility is assigned when a system fails.
Overall, participants learned that an audit needs to begin with a clearly defined risk and follow a documented path through the assessment process. This includes identifying what is being tested, selecting an appropriate methodology, examining evidence, and producing findings that can be explained and compared
Strengthening international cooperation in AI supervision
GNAIS also took part in a panel on the role of international collaboration in AI supervision on 16 September, moderated by Huub Janssen and bringing together GNAIS representatives from Cuba, Luxembourg, Saudi Arabia, and Thailand. AI systems and their supply chains often cross-national borders, while incidents can affect several jurisdictions at the same time. In this context, panellists identified the sharing of knowledge, methodologies, and experiences as a tool that helps authorities respond to common challenges in AI supervision.
The panel also highlighted the importance of developing common approaches and building trust between supervisory authorities and market actors. Greater consistency can support cooperation between jurisdictions and help avoid unnecessary duplication of work, while allowing authorities to continue applying their respective national legal frameworks.
Capacity building was another key theme, particularly for countries that are still developing their AI regulatory and supervisory frameworks. Participants also stressed the value of sharing both successful experiences and lessons learned.
As GNAIS enters its second year, the discussions in Riyadh reflected a growing focus on moving from high-level exchanges towards practical cooperation. The meeting reflected on the Network’s work during its first year of operation and discussed the priorities for the next phase. The priorities highlighted by members include information sharing, international cooperation, capacity building, and greater alignment and harmonisation of approaches.
As the Network continues to grow, its Co-Chair, the Saudi Data & AI Authority (SDAIA), highlighted both the diversity of the Network and the opportunity to expand its membership:
We now have more than 30 members, representing a diverse group of countries, and we invite interested UNESCO Member States to join the Network and take part in the discussions.
Raneem AlkattanGeneral Manager of AI Governance, SDAIA
Overall, the Forum allowed GNAIS members to continue building connections and advance the Network’s role as a platform for international cooperation among AI supervision authorities.
