From Data Spaces to Federated AI: Governing Trusted Intelligence Across Organisations

AI will need learn from data held across organisational, technical and geographical boundaries. AI initiatives do not scale because data is sensitive, ownership is fragmented, policies differ and moving and centralising the data is costly and an increasing security risk.

In this session, Hiro presents its fully decentralised data space approach as a practical foundation for federated AI: organisations keep control over their data products, policies and infrastructure while making selected data and AI capabilities discoverable, governable and usable across trusted ecosystems. Rather than treating a data space as a catalogue or a one-off data-sharing project, we see it as an operational layer for trusted collaboration connecting data products, participants, usage policies, compute capacity and AI services.

In healthcare, one of the most demanding domains, we are building such a data space for the Nextgen project. Clinical, operational and research data must remain under appropriate governance, while providers, researchers, technology partners and public organisations still need to collaborate. Similar challenges appears in energy, manufacturing, mobility, smart cities and other critical sectors.

We will discuss how federated AI can be developed and deployed across a data space, with data remaining close to its source and governed by its owner. The session will also look ahead to GenAI. As organisations introduce retrieval-augmented generation, domain-specific models and agentic services, the need for governed access to trusted data products becomes even more important. A data space can provide the policy, provenance and access-control foundation that allows GenAI to use enterprise and ecosystem knowledge without turning that knowledge into an uncontrolled external dependency.

Attendees will leave with a practical view of how data spaces, federated AI and edge-cloud infrastructure can work together to create scalable, governed AI ecosystems starting in one domain and extending across sectors.

What the audience should take away

• A data space is not merely a sharing portal: it is a governance and operating layer for data products, access rules, participants and AI services.
• Federated AI enables collaboration without assuming that sensitive data must be centrally copied, pooled or surrendered to a platform provider.
• Healthcare is an excellent proving ground, but the pattern translates directly to industrial, energy, public-sector and critical-infrastructure ecosystems.
• GenAI raises the value and the risk of connecting AI to enterprise knowledge; trusted retrieval, policy enforcement, provenance and controlled access become essential.
• Edge-cloud infrastructure can provide the local compute and operational control needed when data, inference or critical processes cannot simply be moved elsewhere.