💰 Funding Details

Overview of the Call

Call Identifier: HORIZON-INFRA-2025-01-EOSC-05

Title: "Using Generative AI (GenAI4EU) for Scientific Research via EOSC"

Type of Action: HORIZON-RIA (Lump-Sum)

Max. EU Contribution: €45 million per project

Submission Scheme: Single stage

Opening Date: 06 May 2025

Deadline: 18 September 2025, 17:00 Brussels time

Strategic Rationale

The call supports large-scale actions that

* embed Generative AI (GenAI) services into the European Open Science Cloud (EOSC),

* create high-quality, machine-readable datasets that comply with FAIR principles,

* enable cross-domain data pooling (health, climate, manufacturing, etc.) under trusted-data governance, and

* run community engagement & training programmes that raise researchers’ readiness to adopt GenAI.

Funding Specifics

* Lump-sum model: The work plan, budget and deliverables are fixed ex-ante; payments are linked to milestone acceptance rather than cost reporting.

* Mandatory Open Science clauses:

* Royalty-free access to results for the EOSC Association (monitoring & policy).

* Deposition of all digital research data in trusted EOSC-federated repositories.

* Non-commercial access rights for EU-identified legal entities to further develop EOSC.

* Partnership obligations: Successful consortia will sign collaboration agreements with projects funded under topic EOSC-03 (AI-readiness & machine-actionability).

Expected Outcomes

1. AI-ready FAIR datasets discoverable and reusable by humans *and* machines.

2. GenAI pilots that demonstrate added scientific value in real-life use cases (e.g. synthetic microscopy images, anomaly detection in environmental sensors, accelerated materials discovery).

3. Protocols, policies & tooling that automate trusted data workflows across the EOSC federation.

4. Scalable training & engagement programmes that mainstream GenAI adoption throughout the EOSC community.

Eligibility Snapshot

* Consortium: Minimum 3 independent legal entities from 3 different Member States or Associated Countries; inclusion of the JRC is encouraged.

* Geographical scope: Entities from your country are fully eligible if your country is a Member State or Associated Country.

📊 Evaluation Criteria

How Your Proposal Will Be Evaluated

1. Excellence (Threshold 4/5)

* Credibility of GenAI methodology: Choice of architectures, bias-mitigation techniques, reproducibility plan.

* Scientific ambition: Ability to “lift science beyond the human scale” through machine-actionable data.

* Interdisciplinarity: Integration of datasets across multiple domains via FAIR protocols.

2. Impact (Threshold 4/5)

* Alignment with EOSC Partnership KPIs (number of FAIR datasets, active users, cross-domain workflows).

* Policy contribution: Concrete support to GenAI4EU, Apply AI strategy and EU Data Strategy.

* Synergy & coordination: Resources devoted to collaborate with EOSC-03 and EOSC governance bodies.

* Sustainability & scalability: Business/operating model for long-term GenAI service provision.

3. Quality & Efficiency of Implementation (Threshold 3/5)

* Work plan robustness under the lump-sum scheme; risk management for AI model drift, data-licensing conflicts.

* Consortium competence & complementarity.

* Gender & diversity measures both in staffing and AI bias controls.

📝 Application Strategy

Winning Application Strategy

1. Build a Mission-Driven Consortium

* Combine domain science leaders (biology, materials, climate…) with AI/ML institutes and EOSC Core providers.

* Leverage your country's excellence in trusted data spaces and industrial AI hubs.

* Include partners from past or ongoing projects (AI4EOSC, iMagine, Data Commons, RI-SCALE) to maximise continuity.

2. Co-design High-Impact Use Cases

1. Map consortium strengths to at least three flagship pilots (e.g. GenAI for pathology imaging, time-series anomaly detection, molecular generation).

2. Quantify scientific KPIs (accuracy gain, time saved, data FAIRness uplift).

3. Align every pilot with the GenAI4EU and Apply AI strategies to demonstrate EU policy relevance.

3. Engineer an EOSC-Native Architecture

* Adopt federated data-mesh patterns (policy-based access, provenance tracking).

* Provide model cards & dataset cards published in EOSC repositories for transparency.

* Integrate AI safety layers (bias detection, explainability dashboards) to meet trustworthy-AI requirements.

4. Exploit the Lump-Sum Model

* Define clear work packages linked to verifiable deliverables (TRL checkpoints, dataset releases, training cohorts).

* Allocate contingency tasks within the lump sum to absorb AI model iteration cycles.

5. Maximise Impact & Sustainability

* Commit to royalty-free EOSC reuse and open-source licensing (Apache-2.0 / EPL-2.0).

* Draft a post-grant service charter describing how GenAI services remain operational within EOSC Marketplace.

* Engage your country National Contact Point early for policy alignment guidance.

💡 Success Factors

EOSC-Native FAIR Compliance

Implement **automated FAIR validators** and publish data/model artefacts with PIDs in EOSC-aligned repositories from day one.

Demonstrable Trustworthy-AI Framework

Embed **bias audits, explainability reports & model cards** as integral deliverables, showcasing leadership in ethical GenAI.

Cross-Domain Flagship Pilots

Select at least **three pilots spanning distinct domains** (e.g. health, climate, materials) to evidence scalability.

Robust Stakeholder Training Pipeline

Deliver **MOOC-style courses, hackathons, and hands-on clinics** in collaboration with your country's research infrastructures.

Active Policy Engagement

Assign a dedicated **policy liaison team** to feed project insights into GenAI4EU and EOSC Partnership boards.

Clear Exploitation & Sustainability Plan

Draft an **open-source governance model** and service-level agreements to keep GenAI tools operational post-grant.

❌ Common Mistakes to Avoid

Ignoring EOSC-Specific Access Rights

Failing to explain **how royalty-free access for EOSC Association and non-commercial users will be operationalised**, leading to evaluator doubts about legal compliance.

Treating GenAI as a Black Box

Submitting technical work packages that *only* mention popular models (e.g. GPT-4) without detailing **bias-mitigation, dataset lineage, or reproducibility protocols**.

Overlooking Training & Community Uptake

Allocating <5 % of the budget to training, despite the call’s emphasis on **awareness-raising and capacity-building programmes**.

Weak Alignment with Existing EOSC Projects

Not referencing outcomes of AI4EOSC, iMagine, RI-SCALE, etc., which evaluators interpret as **duplication of effort**.

Poor Lump-Sum Justification

Providing generic cost tables instead of **deliverable-linked lump-sum breakdowns**, risking negative scores under Implementation.

Single-Domain Focus

Designing pilots limited to one scientific field, thereby **undermining the cross-domain data pooling objective**.

Using Generative AI (GenAI4EU ) for Scientific Research via EOSC

Quick Facts for Spain Applicants

Ready to Apply from Spain?