The third edition of the European Data Conference on Reference Data and Semantics (ENDORSE) brought together in Brussels, on 8 and 9 October 2025, the semantic interoperability community with 78 speakers – including three keynote speakers – more than 250 onsite participants, and almost 700 online views via web streaming.
Organised by the Publications Office in collaboration with Interoperable Europe, ENDORSE 2025 focused on how reference data, multilingual semantic assets, and trustworthy AI power the next generation of digital public services and shape Europe’s digital future.
The Director-General of the Publications Office of the European Union and the European Commissioner for Trade and Economic Security; Interinstitutional Relations and Transparency opened ENDORSE 2025 with speeches setting the tone for high-level exchanges and hands-on demonstrations. The conference provided a platform for experts from EU institutions, Member States, academia, and SMEs to explore key themes such as:
- The semantic foundations of AI readiness and responsible automation in the public sector
- The role of reference data in Common European Data Spaces
- The practical use of tools like VocBench, ShowVoc, and the new SELEN platform
- The high interest in data stewardship, AI agents, and data labs as enablers of governance and innovation
One of the core takeaways?
High-quality, multilingual, and structured reference data is clearly recognised as a cornerstone of trustworthy AI and next-generation public services. Grounding AI systems in standardised vocabularies, classifications, and ontologies reduces hallucinations, improves explainability, and enhances reliability, especially in multilingual and cross-border contexts.
This year’s edition also brought policy into sharp focus, aligning closely with the AI Act, the Interoperable Europe Act, the Data Governance Act, and the wider European Data Strategy, which together are shaping a future where data is interoperable, multilingual, and reusable by design.
What did participants highlight in their feedback?
- The clear need for shared semantics and common understanding to enable true interoperability.
- Knowledge graphs, ontologies, and neuro-symbolic AI were frequently cited as key building blocks for intelligent, trustworthy systems.
- Enthusiasm around emerging synergies between LLMs and the Semantic Web, while acknowledging technical challenges in tools like SHACL and RML.
- An urgent call for ensuring good data quality, curation, and strong data stewardship to ensure reliable AI and machine-readable government.
- Emphasis on the human factor, the role of data stewards in bridging data producers, domain experts, and AI specialists.
- A shared understanding that agreement on reference data is essential, painstaking, and foundational to all efforts.
- A strong desire to work together, across systems and communities of practice, uniting AI developers, semantic experts, and public administrations.
These insights confirm that semantic interoperability is a shared strategic imperative. ENDORSE 2025 has once again proven that progress lies in collaboration, and that the semantic layer is where policy, people, and technology meet.
Learn more and explore the programme: https://op.europa.eu/en/web/endorse-2025/programme
Photos by Daria Prykolota