At the EOSC Symposium 2025 in Brussels, nearly 500 participants from 36 countrieswitnessed a pivotal moment: the official launch of the EOSC Federation.The coordinators of 14 EOSC Nodes of the EOSC Federation joined President Klaus Tochtermann of the EOSC Association on stage to sign the Federation's Memorandum of Understanding amid confetti and fanfare. This festive moment marked took a decisive step for Europe's research infrastructural landscape: from vision to operational reality.
EC Director-General Marc Lemaître captured the momentum: "Let's seize this moment! Let's act together to complete the EOSC Federation fast and make it the backbone of Europe's digital research and innovation ecosystem."
FAIR for AI: Consistency as the foundation
The symposium's "FAIR for AI" session brought together diverse perspectives on how FAIR principles intersect with AI systems, from technical interoperability to metadata standards and the evolving needs of AI researchers.
As Co-Chair of the EOSC FAIR Metrics and Digital Objects Task Force alongside Mark Wilkinson (Universidad Politécnica de Madrid), Elli Papadopoulou (Athena ResearchCenter) brought the conversation to a critical challenge: assessment consistency. We highlighted how assessment tools interpret FAIR principles differently, measure different aspects, and produce incompatible metrics.If two research teams can evaluate the same dataset and get wildly different scores (e.g. one at 85%, another at 62%)this can create a potential trust problem for the EOSC Federation.
The response from attendees was immediate and strong. Many shared their own experiences with inconsistent results and confusion around what "FAIR" actually means in practice. The Task Force has documented this through community surveys, and the session confirmed it resonates deeply across the research community.
OSTrails addresses this directly through the FAIR Interoperability Framework (FAIR-IF) and the FAIRness Reference Model, work that received specific attention because it tackles a gap everyone recognises. As the EOSC Federation is built up from multiple nodes needing to evaluate and exchange data, standardised interoperability and assessment frameworks and practices become foundational.
In the discussion also a harder truth surfaced. A dataset can score perfectly on FAIR metrics and still be biased. The dataset can have DOI, open license, standard formats, and rich metadata, but still carry information from limited demographic groups or lacking documentation of curation decisions and accountability trails which limits its AI readiness.
FAIR principles ensure technical accessibility and reusability. AI systems require additional dimensions like bias assessment, curation quality documentation, ethical context metadata, and accountability frameworks (Microsoft has labeled thisthe FATE principles: Fairness, Accountability, Transparency, Ethics). OSTrails ensures the FAIR part is reliable and consistent, so that where FAIR and FATE intersect, there's a trustworthy foundation to build on.
Standards in action
The lightning talk sessions transformed principles into practice. Jakub Jirka from Codevence demonstrated their DMP platform, including a first implementation of the DMP evaluation service that Athena Research Center and TU Wien are co-developing as part of OSTrails. The demonstration showed machine-actionable workflows that help researchers achieve FAIR compliance while addressing quality, bias, and ethical dimensions.
What’s Coming
The EOSC Federation launched with 14 nodes, with more nodes on the way. OSTrails is timely entering its implementation phase where frameworks are becoming operational reality.
Over the coming months, the project pilots will demonstrate how the OSTrailsinteroperability standards work across national, institutional, and domain-specific infrastructures, some of which are already building their Node to be part of the Federation. Stay Tuned!
