Collaborative session notes:
Open session notes
Description
Artificial intelligence is now embedded in every layer of the research data ecosystem, from data preparation and curation to federated access, visiting-data models, and automated decision-making in secure processing environments. As systems move rapidly from prototype to production, the need for clear, actionable AI governance frameworks has become central to research data resilience. This session examines how ethics, governance, and technical implementation intersect to support trustworthy AI systems capable of interacting with sensitive, distributed, and policy-constrained datasets.
Drawing from recent work of the EOSC-Future/RDA Artificial Intelligence and Data Visitation Working Group (AIDV-WG), the discussion brings together policy, ethics, and technical insights from: the DV4RDA initiative and the FAIRlyz platform for biomedical data sharing and analysis; guidance for ethics committees and informed consent in AI and data visitation; emerging legal foundations for secure processing environments; federated and visiting-data architectures explored in ‘Geographies of Trust’; and the AIDV-WG’s shared citation library documenting global developments in AI oversight. The session focuses on how AI governance can strengthen resilience by ensuring continuity of access, maintaining trust in data flows, and enabling responsible automation across research infrastructures.
Participants will explore how governance frameworks translate into operational tools, how federated and visiting-data models challenge established ethics and compliance practices, and how communities can coordinate governance approaches that both enable innovation and protect sensitive data. The session encourages dialogue on developing shared RDA principles for AI governance in support of resilient, trustworthy, and interoperable research data ecosystems.
Principal Objective
To explore how ethics-aligned AI governance frameworks (supported by practical tools, platforms, and legal foundations) can enhance research data resilience and continuity, and to assess how the Research Data Alliance can support the further development of guidance and tools in this domain.
Supporting Objectives
1. Examine how federated and visiting-data models reshape governance requirements by drawing on insights from the AIDV-WG’s Geographies of Trust report, including the challenges of cross-jurisdictional data flows, distributed responsibilities, and new trust relationships between humans, systems, and autonomous agents.
2. Showcase practical examples of ethics-aligned tools and platforms (including the DV4RDA project and FAIRlyz) that demonstrate how governance, secure access, and AI-assisted analysis can be operationalized for biomedical and other sensitive research data domains.
3. Bridge ethics and implementation by examining how frameworks such as the AIDV-WG’s Guidance for Ethics Committees Reviewing AI and Data Visitation, Guidance for Informed Consent, and the AI Bill of Rights Recommendations can be translated into actionable practices, workflows, and risk controls during system deployment.
4. Explore pathways for moving AI systems from prototype to production based on insights from the Kaggle ‘Prototype to Production’ report, with attention to reproducibility, safety, reliability, and the governance mechanisms needed to support long-term continuity.
5. Discuss legal and policy foundations coherent governance models that allow computation to move to the data, preserve privacy, and maintain research continuity during disruptions.
6. Consider how shared knowledge resources, such as the AIDV-WG Shared Citation Library, can support community learning, harmonisation of governance practices, and development of interoperable approaches across global research infrastructures.
7. Identify RDA community priorities interests in the development AI governance aimed at supporting coordinated development of guidelines, policy recommendations, digital instruments.
