Collaboration between european data spaces: 131 participants, 9 sectors, and one shared roadmap

On June 25, 2026, Visions and Prometheus-X hosted a webinar on collaboration between European data spaces: open source, interoperability, and what it actually takes to make data spaces work together.

131 participants joined from across Europe: sectoral data spaces, enterprises, researchers, and representatives from the European Commission.

Here’s what was presented, what came out of it, and what happens next.

The starting point: A demanding model, a fragmented reality

Matthias De Bièvre (CEO, Visions / President, Prometheus-X) opened with the core tension shaping the whole sector:

Europe’s data and AI economy needs to meet four requirements at once:

• Pool and protect private data, without organisations losing control of it
• No lock-in: no single cloud provider or dominant platform capturing the value
• Independent audits: algorithms must be auditable by independent third parties
• Trust in usage: confidence that shared data is used only as agreed

Data spaces are the answer to all four. But today, each data space is its own island. No shared technical stack, no common governance model, no aligned business logic. That combination is what blocks the emergence of one interoperable network.

The response: open source + interoperability.

• Pool shared building blocks to accelerate development for everyone
• Build interoperability that goes beyond protocols, covering governance and business models too

That double thread ran through the entire session. Part 1 showed pooling in action across five live use cases. Part 2 opened up the harder question: what happens when data spaces actually need to work together.

Part 1: Five data spaces, five use cases, one shared infrastructure

Each speaker followed the same structure: their use case, what they already use from the Prometheus-X open-source stack, and what they contribute back. That two-way reading, use and contribute, is what gives the commons its value. Every brick one sector adds becomes immediately available to all the others.

Health DataTrust: Turning medication waste into a secure exchange

Mariane Cimino (Director of Operations, Health DataTrust) presented a use case with real economic weight:

• €500M to €1.7B in medication wasted every year in France
• 1/3 of that waste is due to expiry alone

Health DataTrust connects hospitals with surplus stock to hospitals that need it, acting as a regulated intermediary that never stores the data itself.

The technical core: data processing stays local, inside each hospital’s own data warehouse. Health DataTrust operates as a structured compliance layer for the EU Health Data Space (EHDS), built around strict compute-to-data and secure processing environment requirements.

On the question of sharing patient data across legally distinct organisations, Mariane clarified that EHDS obliges every data provider to expose its data to any requesting user, under defined consent and security conditions, and that its primary-use chapter mandates interoperability with the EU’s MyHealth@EU infrastructure.

Contributed back to Prometheus-X:

• Sector-specific regulated rules in machine-readable format (compute-to-data obligations, anonymisation requirements)
• A metadata extraction and visualisation component for populating the data catalogue

Legal Data Space: AI agents that draft data contracts

Martin Bussy (Co-CEO, Legal Data Space) showed how connecting private contractual data with public regulatory sources lets an AI agent read usage context and generate a valid, enforceable data contract automatically.

The goal: a law-as-a-service layer, applicable across all data spaces, guaranteeing compliance, traceability, and contractualised data usage.

Architecture: an AI agent orchestrator connects to VisionsTrust via ALIEN, dynamically deciding which data contract to activate based on context. Services go live October to November 2026.

Contributed back to Prometheus-X:

• An extended contract ontology covering legal basis, data sensitivity, and compute-to-data obligations
• A reusable ODRL policy library, covering AI usage granularity (index, RAG, fine-tuning, distillation), human-in-the-loop requirements, mandatory attribution, and transfer restrictions
• Available in the Prometheus-X ODRL Policy Registry: October 2026

LIST (Luxembourg): Matching skills to the job market

Marie Gallais and Marcos Da Silveira (Luxembourg Institute of Science and Technology) presented a skills intelligence service connecting employment agency data (ADEM) with training catalogues from Luxembourg training providers. It extracts skills from job postings via ESCO-based annotation, combining generative AI with semantic graph mapping.

Live demo results:

• 668 documents, 17,555 skills extracted
• 96.2% human acceptance rate
• versus 93.3% for an LLM-only approach

Data exchange runs through Prometheus-X connectors (PDC), with contracts and consent managed via VisionsTrust.

Contributed back to Prometheus-X:

• A skills data structure extending both ELM and ESCO, a building block any data space can reuse to represent skills interoperably

AFP / Tralalère: The first live bridge between two different protocols

Isabelle Wirth (AFP) and Sylvie Tissot (Tralalère) presented arguably the most significant technical demonstration of the session: fighting misinformation as a real, production cross-data-space use case.

AFP’s fact-checking content, exposed via the TEMS media data space (EDC protocol, operated by Startin’blox), is discovered and contracted from the DS4Skills education data space (Prometheus-X, operated by VisionsTrust). Two ecosystems running completely different tech stacks.

What it enables:

• An AI debunking assistant helping learners identify and counter misinformation using AFP-verified fact-checks
• An educational video generator producing short debunking videos for social media

How the bridge works: the Dataspace Protocol’s (DSP) contract negotiation protocol, a direct trust configuration between TEMS and DS4Skills, and the DSIF’s root and sector authorities enabling cross-catalogue discovery between ecosystems.

This is the signal worth noting: cross-protocol interoperability is no longer a hypothesis, it’s a bridge running in production today. The code is already public on GitHub, built as a reusable pattern for future PTX to other-ecosystem bridges.

Techné: Culture as an entry point into other sectors

Fayrouze Masmi-Dazi presented Techné, the French cultural data space connecting media operators and cultural institutions to fix today’s fragmentation of audience, agenda, and ticketing data. Priority use cases: audience extension, data-driven revenue sharing, ticketing, and cultural agenda visibility.

The most interesting part for the wider ecosystem: Techné showed interconnection that runs both ways.

• A pedagogical assistant connected to an education data space can recommend cultural events matching a learner’s profile, and Techné’s agenda data can in turn inform which events best fit which learning paths
• A travel assistant connected to a tourism data space can recommend cultural events based on a traveller’s trip and interests, and the tourism data space can recommend destinations and trips to Techné’s cultural audiences based on their interests

A two-way recommendation loop, not one-way distribution.

Part 2: What if data spaces actually had to work together?

Part 2 shifted register entirely. Instead of sectoral demos, a roundtable, Lydia Montandon (AnySolution / DeployTour), Jonathan Huffstutler (EONA-X), Antti Jakobsson (Location Data Space), and Thimo Thoeye (OASC / DS4SSCC), tackled a deliberately demanding scenario: a red heat alert in Paris.

The scenario: A tuesday in july, 38.3°C, and a major landmark closure

• Red alert across 16 regions
• 1,000+ schools closed as a precaution
• A major Paris landmark closes its summit to visitors without pre-purchased tickets

The closure is the right call, and it happens regardless. But the crowd doesn’t disappear: it moves. And no single data space, alone, has the visibility to anticipate what happens next.

Four layers, combined, change the picture:

• Location Data Space: building footprint, height, and shade-mapping data; available capacity in nearby indoor sites
• DS4SSCC: the BeatTheHeat methodology (already deployed in Cartagena, Naples, and Taranto), combining forecasts, building characteristics, and demographic data to flag urban heat islands up to 72 hours ahead. Presented here as a model transferable to Paris, not as a system already running there.
• DeployTour: destination intelligence combining ticket cancellations and visitor behaviour to anticipate crowd displacement toward nearby gardens and museums
• EONA-X: targeting transport heat protocols at the specific stations under combined thermal and crowd pressure, instead of applying the same measures network-wide

The honest framing: this isn’t a system that prevents the closure. It happens anyway, and it’s the right call. What changes is lead time and targeting.

• Without Location Data Space: no building-level view of which structures could absorb the redirected crowd
• Without DS4SSCC: the closure decision is made same-day, with no 72-hour anticipation
• Without DeployTour: the secondary crowd displacement isn’t seen coming
• Without EONA-X: transport heat protocols still activate (they always do), just uniformly, not targeted

What’s real today, what isn’t yet

Each panelist gave a candid breakdown, covering data, technical infrastructure, and governance, of what’s actually operational versus still in progress:

• Antti Jakobsson (Location Data Space): already connects high-value datasets across European countries, including IGN France building data, computing 3D data on the fly via a Finnish supercomputing environment. Currently OGC-API-based rather than a full unified data structure; an MVP connecting multiple data spaces (building capacities, forestry data) is expected later this year.
• Lydia Montandon (DeployTour): the goal isn’t real-time, it’s near-real-time. The main challenge isn’t technical but lies in data-provider agreements and federating existing spaces. Operational with connectors, but governance rules enabling tourism operators to benefit from other data spaces’ signals are still being built.
• Thimo Thoeye (DS4SSCC): a complex federated project, 11 pilots, 26 cities, 50+ partners today, targeting 10 more pilots and 20 more cities. Faces real technical fragmentation across EDC, FIWARE, and distributed ledger stacks; collaborating with the EU’s Simpl programme on a federation intermediary.
• Jonathan Huffstutler (EONA-X): an operational entity with 25 full-time staff, a scientific board, and a board of directors, actively engaged in federations like DeployTour and the EU’s Common European Mobility Data Space. His key point: no rule today mandates interoperability between European data spaces. Collaboration depends on business-driven use cases and voluntary alignment.

DSIF: Where the technical and governance work actually stands

This is the context the Data Space Interoperability Framework (DSIF) update spoke to directly, with a clear line between what’s done and what’s in progress.

Technical infrastructure

Already in production:

• Catalogue federation via DSIF root and sector authority components, integrated with Prometheus-X, EDC, FIWARE, Athumi/Solid, live across the Skills and Media data spaces
• Cross-ecosystem contract negotiation between the Prometheus-X connector and TEMS’ EDC connector, via DSP’s Contract Negotiation Protocol: the exact mechanism behind the AFP/Tralalère bridge above
• FIWARE connector augmentation with Data Space Protocol to TM Forum mapping, available for the smart cities data space

 In progress:

• Contract model mapping across PTX, EDC, deltaDAO, and FIWARE, first implementation on smart cities cases, October 2026
• DSIF consent specification with PTX, EDC, and Solid integrations, first implementation on health and skills cases, October 2026

Governance: the Rulebook & policy commons

Olivier Dion presented DSIF’s governance work as a shared baseline of roles, rules, and reusable verification evidence across data spaces.

• 124 rules in the DS4Skills rulebook, now the seed for DSIF alignment across data spaces
• Goal: portability, not a one-size-fits-all rulebook

Five common roles defined so governance can be compared and verified across data spaces:

• Governance Authority
• Data provider
• Service / AI provider
• Data intermediary
• Certificator

Four deliverables:

1. Rulebook: roles, rules, conformance expectations
2. Policies: machine-readable (ODRL) and human-readable
3. Ontology: shared contract concepts across data-space stacks
4. Methodology: reusable role/rule/evidence mapping approach

Timeline:

• • End of July 2026: DS4Skills Rulebook v1 plus first feedback from other data spaces
  • September to December 2026: testing with first data spaces and cross-data-space use cases
  • January 2027: final common Rulebook plus contract templates, policies, and ontology

What the Q&A surfaced

A central tension, raised by Alex Tourski: why keep multiplying sectoral data spaces, when for an end user like a farmer, agriculture, tourism, and logistics are inherently linked? 

Matthias De Bièvre acknowledged it directly: the Commission’s vision is a single data market, but sectoral data spaces launched first to enable within-sector exchange, which is exactly the fragmentation interoperability work now has to resolve. Corentin Sastre added in chat that governance questions, who decides, how, who’s accountable for what, need solving before the technical ones.

Recurring technical blockers identified for combining data across spaces:

• Catalogue discovery
• Contract models
• Protocols
• Identity management
• Audit procedures

On metadata federation, Sten-Erik Björling raised the question of standardised methodologies for mapping dataset content, strengths, and limitations for AI model evaluation. Thimo Thoeye confirmed the DS4SSCC Blueprint already addresses several metadata building blocks, with DCAT-AP as the common baseline, though many pilots run their own extensions. Marie Coussement noted that the EU’s Simpl programme is currently scoring trusted open-source technologies for its Simpl-Open AI/ML extensions, with details coming soon via their LinkedIn group.

Other threads covered: data space certification against DORA/NIS2, the Legal Data Space’s potential role as a rulebook-as-a-service intermediary, and whether LIST’s ESCO-based annotation service is available for other Skills Dataspace actors to test.

Next steps with the community

Three concrete invitations closed the session:

• Fill in the Mural. Throughout the session, a collaborative Mural let participants propose data, services, use cases, or expertise relevant to the projects presented. Submitted contributions will be reviewed, and contributors will be contacted directly to discuss involvement.
• Join future workshops. Mural contributions will directly feed upcoming workshops, with announcements to follow.
• Join the DSIF Rulebook Working Group. The most direct call to action: any organisation wanting to help design the common rulebook for data space infrastructure can join the working group by emailing Olivier Dion, olivier@onecub.fr. Several participants, including from energy, transport, and training data spaces, already expressed interest live during the session. To work on the technical interoperability, participants can contact Félix Bole : felix@visionspol.eu

The takeaway

The real signal from this session is the consistency between its two halves.

Part 1 showed that open-source pooling already works in practice: five very different data spaces (health, legal, skills, media, culture) run on the same shared infrastructure and contribute back to it, each one strengthening the whole.

Part 2 showed, with real candour about the gaps, that going further, getting multiple data spaces to collaborate on a real scenario like an urban heatwave, is less a technical challenge than a shared governance one.

That’s exactly where DSIF is working now: a timeline, identified deliverables, and an open invitation to contribute.

The next step won’t happen in a conference room. It’ll happen in the iterative work of the rulebook working group. If your organisation runs a sectoral data space, or has data, services, or use cases to bring to this ecosystem, the Mural is still open, and so is the working group.

Thank you to the 131 participants who joined this session, to our speakers, Mariane Cimino (Health DataTrust), Martin Bussy (Legal Data Space), Marie Gallais & Marcos Da Silveira (LIST), Isabelle Wirth (AFP) & Sylvie Tissot (Tralalère), Fayrouze Masmi-Dazi (Techné), Lydia Montandon (AnySolution/DeployTour), Jonathan Huffstutler (EONA-X), Antti Jakobsson (Location Data Space), Thimo Thoeye (OASC/DS4SSCC), and Olivier Dion (DSIF Rulebook), and to everyone who contributed live via chat and Mural.