Session date and time: Breakout Session 1 – 12:00-13:30 BST, Tuesday, 6 October 2026

Collaborative session notes

The EOSC-Future/RDA Artificial Intelligence and Data Visitation Working Group (AIDV-WG) is an existing RDA group with endorsed outputs and a recognised maintenance role. The purpose of this P27 session is to advance the interests of the AIDV-WG by focusing on implementation of its outputs, adoption of data visitation practices, responsible use of AI in controlled-access data environments, maintenance of endorsed recommendations, and the transition pathway towards a new Interest Group and Working Group. 

This session will outline the transition of the AIDV Working Group (WG) into a new AIDV Interest Group (AIDV-IG) and the official chartering of the Visiting Agent Network & Trusted Access Governance Evaluation WG (VANTAGE-WG). This new working group is designed to bridge the gap between technical communities developing agentic AI and the governance bodies overseeing ethics, sovereignty, data protection, and research integrity. Its central purpose is to build a shared RDA agenda in which agentic AI systems and governance are developed together, rather than treated as separate domains.

Agentic AI is emerging as a major new capability in research data systems, open science infrastructures, scientific workflows, and global open research commons. Agentic AI systems can plan, act, retrieve information, interact with software tools, query data environments, initiate workflows, make recommendations, support research operations, and coordinate complex tasks across distributed data, software, infrastructure, and governance environments. These capabilities create significant opportunities for scientific discovery, data stewardship, research automation, responsible data reuse, and cross-disciplinary collaboration. They also raise important questions about ethics, law, sovereignty, human rights, accountability, institutional responsibility, system safety, auditability, reproducibility, and the relationship between human judgement and autonomous or semi-autonomous computational action.

The session will convene existing and new RDA members to examine how AI and agentic AI can be implemented via data visitation across global open research commons, especially where research relies on sensitive, distributed, controlled-access, or jurisdictionally complex data. Data visitation is central to this discussion because it offers a practical pathway for open science in which data do not need to be moved, downloaded, copied, or centralised before responsible analysis can occur. In doing so, the session will connect the core principles of AI Data Visitation (AIDV) with emerging developments in AI, clarifying how these approaches can strengthen the next generation of research infrastructures and global data commons.

As part of its introduction to VANTAGE‑WG, the session will situate agentic AI within the broader landscape of emerging technologies, including digital twins, federated analytics, privacy‑preserving machine learning, secure and trusted research environments, autonomous scientific instruments, robotics, synthetic data systems, knowledge graphs, and AI‑enabled research assessment tools. These technologies are increasingly interconnected. Agentic AI may orchestrate them, act within them, or depend on them. For this reason, the session will address agentic AI as part of a broader socio-technical ecosystem rather than as a stand-alone technology.

VANTAGE-WG will not treat governance as an external review layer added after technical development. Nor will it treat technical development as a purely engineering activity that can proceed independently of legal, ethical, institutional, and societal responsibilities. Instead, it will explore how agentic AI systems must be designed, programmed, deployed, evaluated, and maintained through reciprocal dialogue between system builders and governance actors.

Supporting Objectives

• Examine how data spaces, including the International Data Spaces Association approach, the Dataspace Protocol, and emerging research data space pilots, can provide machine-actionable governance, policy negotiation, access control, auditability, and interoperability for implementing data visitation across distributed research environments.
• Discuss implementation opportunities emerging in Indonesia, Asia and the Western Pacific, Africa (Kenya, Nigeria), Belarus, Kazakhstan, and other CIS countries, with particular attention to ethics committees, regulatory authorities, ministries of health, health data systems, and regional research governance structures.
• Identify adoption needs across research communities and global initiatives, including health, genomics, environment, crisis data, pharmacovigilance, agriculture, social sciences, cultural heritage, repositories, research assessment, and open research commons.
• Consider how AIDV-WG outputs can contribute to UNESCO-endorsed projects under the United Nations International Decade of Sciences for Sustainable Development, 2024-2033, including DPs4Crises-Pilot, HGP2, and SE4RA. UNESCO states that the UN General Assembly declared 2024 to 2033 as the International Decade of Sciences for Sustainable Development and designated UNESCO to lead its implementation.
• Explore how AIDV-WG and DV4RDA outputs can support the proposed CAI3R-science and AI4GHR centres within the UN ODET-supported Global Network of Centres for Exchange and Cooperation on AI Capacity Building. The UN Office for Digital and Emerging Technologies was established on 1 January 2025 to advance inclusive, rights-based digital and emerging technology cooperation.
• Define next steps for AIDV-WG maintenance, global dissemination, adoption tracking, and the proposed transition to the RDA Artificial Intelligence and Data Visitation Interest Group, while keeping the separate Agentic AI Working Group distinct from the AIDV continuation pathway.
• Address the critical disconnect between the technical potential of autonomous agents in data research and the organizational readiness required to deploy them safely. 
• Bridge the gap between high-level innovation and the practical realities of institutional oversight of open data commons, ensuring that agentic workflows do not compromise security or regulatory compliance.

1. Welcome and Session Purpose: Seonyoung Kim (5 min)
2. AIDV implementation opportunities: GERP, WHO, UN: Francis P. Crawley (10 min)
3. Pathway to long-term coordination: Francis P. Crawley (5 min)
4. Proposal for RDA VANTAGE-WG: Patricia Buendia (5 min)
5. Why Persona-based agentic AI systems matter for the Research Data Lifecycle: Raymond Uzwyshyn (10 min)
6. Speaker presentations by invited contributors (25 mins total – 5 minutes per speaker)
   • Data Governance and security for AI agents: Lars Eklund
   • Managing the Monkey: A Socio-Technical Approach to Human-in-the-Loop and Governance for Agentic AI: Gnana Barathy
   • Institutional data‑access and protection policies governing AI agents: Anthony Juehne
   • Emerging technology intersections: digital twins, federated analytics, privacy-preserving machine learning, autonomous workflows, synthetic data, knowledge graphs, and AI-enabled research assessment: Patricia Buendia
   • From Principles to Practice: A Trust-Centred Governance Framework for Deploying Agentic AI in African Research Ecosystems: Ugochi Okengwu
7. Open Discussion and Q&A: Seonyoung Kim (20 min)
8. Poll and Next Steps: Seonyoung Kim (5min)
9. Closing remark: Francis P. Crawley & Seonyoung Kim (5 min)

The session is intended for agentic AI developers, programmers, research software engineers, system builders, deployers, data scientists, infrastructure providers, data space developers, data stewards, repository managers, ethics committee members, institutional review board members, legal experts, human rights specialists, policymakers, research governance officers, research funders, research performing organisations, librarians, open science practitioners, and representatives of global open research commons.

It will be especially relevant to RDA members working with agentic AI, autonomous or semi-autonomous research workflows, secure processing environments, trusted research environments, sensitive data, controlled access, federated analytics, data spaces, AI-enabled research discovery, research assessment, data governance, and open science infrastructure.