Skip to main content
European Commission logo
Reforms and Investments
AI, Machine learning, Hands of robot and human touching on big data network connection background, Science and artificial intelligence technology

Supporting reforms to reinforce efficient and high quality public administrations

Funding Programme
Year
  • 2023

Supervising Artificial Intelligence by Competent Authorities

The European Commission strengthened the capacity of the Dutch Authority for Digital Infrastructure (RDI) and members of the Dutch and European Working Groups of Competent Authorities on AI to effectively supervise AI. UNESCO provided analysis, targeted capacity-building, practical tools, and multi-stakeholder consultations to support AI supervision in compliance with the EU Artificial Intelligence Act and international ethical standards.

Context

The adoption of the EU AI Act created new requirements for competent authorities supervising AI systems. Authorities needed to develop new skills, use appropriate assessment tools, and strengthen coordination to implement the AI Act and related legislation cohesively. 

At the same time, UNESCO’s Recommendation on the Ethics of AI provides global standards for the ethical development and use of AI. The project brought UNESCO’s expertise and experience in AI governance to support the Dutch Authority for Digital Infrastructure (RDI) and EU competent authorities in strengthening supervisory capacities.

Support to be delivered

From December 2023 to December 2025, UNESCO provided structured support to the Dutch Authority for Digital Infrastructure (RDI) and members of the Dutch and European Working Groups of Competent Authorities on AI. Activities covered three phases: mapping and analysing AI supervisory practices in the EU and beyond; targeted capacity-building through case studies and tools; and multistakeholder consultations to develop operational recommendations. 

The project delivered training sessions, expert roundtables, practical supervision tools, case studies, training materials, and an AI Supervision Training Toolkit. The Dutch Authority for Digital Infrastructure (RDI) was the main beneficiary and key partner.

Results achieved

The project achieved its planned results. More than 700 civil servants from over 15 EU Member States participated in in-person training activities at European and national levels. Eight specialised AI Supervision modules generated more than 200 single registrations. 

The project developed three real-world case studies, two extensive training use-case documents, a Compliance Map, an AI supervision board game and an AI Supervision Training Toolkit. According to participant feedback, 97% of respondents indicated that the training addressed their needs and expectations, while 100% confirmed that case-study exercises enhanced their understanding of key supervisory concepts. The project also contributed to the creation of the Global Network of AI Supervisory Authorities (GNAIS).

Useful links

https://www.unesco.org/en/artificial-intelligence/recommendation-ethics/supervising

Launch event, The Hague, March 2024

GNAIS webpage: https://www.unesco.org/ethics-ai/en/global-network-ai-supervisory-authorities

Global Landscape report executive summary: https://unesdoc.unesco.org/ark:/48223/pf0000393208

Enhancing Governance: Use Cases for Supervisory Authorities

Case Studies

Blog posts:

Material for training sessions

Others

Report: Pathways on Capacity Building for AI Supervisory Authorities: https://unesdoc.unesco.org/ark:/48223/pf0000396637

Web story on the Expert Roundtable series: https://unesco.org/en/articles/unesco-hosts-expert-roundtable-series-ai-supervision-and-capacity-building

Web stories on the training sessions:

Web story on Policy Dialogue event: https://www.unesco.org/en/articles/international-panel-explores-complexities-ai-supervision

RDI news story about the closing event: https://www.rdi.nl/onderwerpen/technologische-ontwikkelingen/kunstmatige-intelligentie/ai-congres/verslag

More about the project

You can read the documents related to the project here: