EU AI in Science Action Plan 2025 | RAISE Initiative, Funding and Research Strategy

The EU Launches a Comprehensive AI in Science Strategy

The European Commission’s AI in Science Action Plan (COM/2025/724) represents the most ambitious institutional attempt to systematically integrate artificial intelligence into scientific research across all disciplines. Published as part of the broader EU AI strategy, this action plan acknowledges a fundamental reality: AI is no longer a niche tool for computer scientists but a transformative technology reshaping how research is conducted across physics, biology, chemistry, social sciences, and every field in between.

The action plan arrives at a critical juncture for European competitiveness. While Europe led the world in AI scientific publications until 2017, both China and the United States have since overtaken the EU, with China now leading globally in AI research output. Europe’s share of global AI players stands at just 6%, and its share of AI patents at a mere 3%. These figures signal an urgent need for coordinated action to prevent Europe from falling further behind in the AI-driven transformation of scientific research.

At the heart of the strategy is a dual objective: advancing “AI in science” (deploying AI tools across all research domains) and “science for AI” (conducting cutting-edge fundamental AI research). This bidirectional approach recognizes that scientific breakthroughs increasingly depend on AI capabilities, while AI itself advances through scientific insights. For research institutions and organizations tracking major policy developments, this action plan will shape European research funding and infrastructure for the coming decade.

RAISE: A Virtual European Institute for AI Research

The flagship initiative of the action plan is RAISE — Resource for AI Science in Europe. Conceived as a virtual European institute, RAISE will pool talent, computing resources, data, and research funding across the EU to create a coordinated approach to AI in science that no single member state could achieve alone. The initial pilot launches in Q4 2025 at the AI in Science Summit in Copenhagen, funded with EUR 108 million from the Horizon Europe 2026-27 Work Programme.

RAISE is built on two pillars. The first comprises Thematic Networks of Excellence in specific scientific disciplines — bringing together the best researchers working at the intersection of AI and domain sciences. The second is a European Network of Frontier AI Labs focused on advancing the fundamental capabilities of AI itself. Together, these networks are designed to address Europe’s key weakness: the fragmentation of research efforts across 27 member states.

The governance structure includes representation from both the AI research community and domain scientists, member states through the AI Board, private sector partners, and a high-level academic advisory board. This multi-stakeholder approach reflects the Commission’s recognition that effective AI in science requires collaboration across traditional institutional boundaries. The phased implementation — pilot under Horizon Europe, then scaling under the new Multiannual Financial Framework — provides a path from concept to continental-scale impact.

EUR 8 Billion in Horizon Europe AI Funding

The financial commitment behind the action plan is substantial. Horizon Europe has supported AI with more than EUR 8 billion since 2021, comprising EUR 6.4 billion for the 2021-2024 period and an additional EUR 1.6 billion in the 2025 Work Programme alone. Of this latest allocation, approximately EUR 700 million is directed specifically to AI in science applications.

The funding flows through multiple channels. The European Research Council (ERC) awarded approximately EUR 450 million in AI-related grants in 2023, with cumulative AI research funding exceeding EUR 2 billion since 2007. Marie Skłodowska-Curie Actions have funded over 1,000 projects with significant AI components. The European Innovation Council committed EUR 150 million to AI projects in 2024, with EUR 400 million across the 2021-2024 period.

Perhaps most significantly, the Commission has signaled its intention to double Horizon Europe’s yearly investment in AI by 2028, including a specific doubling of AI in science funding. However, the action plan acknowledges a critical structural challenge: approximately 90% of public research funding in the EU comes from the national level. Effective implementation therefore requires not just EU-level investment but coordination with 27 national research funding systems — a governance challenge that will test European institutional capacity.

Closing the Computational Capacity Gap

One of the most striking data points in the action plan is Europe’s computational deficit. The EU holds less than 5% of global AI computational capacity, compared to 75% for the United States and 15% for China. This gap directly limits European researchers’ ability to train large AI models, process massive datasets, and compete in computationally intensive AI research.

The Commission’s response centers on two infrastructure categories. AI Factories are AI-optimized EuroHPC supercomputers enhanced with services specifically designed for research use. AI Gigafactories represent the next level — large-scale facilities integrating massive computing power with energy-efficient data centers capable of supporting the most demanding AI training and deployment workloads. Up to EUR 600 million from Horizon Europe has been allocated for dedicated research access to these facilities.

The plan also emphasizes guaranteed availability and scheduling priority for EU-funded research projects, addressing a common frustration among European researchers who often face long queues and limited access to computational resources. The integration of quantum computing capabilities is mentioned as a forward-looking component, recognizing that the convergence of quantum and AI technologies could reshape computational possibilities within the next decade.

Data Infrastructure and the European Open Science Cloud

AI in science depends on access to high-quality, well-curated datasets. The action plan positions the European Open Science Cloud (EOSC) as the backbone of the EU’s research data infrastructure, serving as the Common European Data Space for Research and Innovation. EOSC provides a federated platform for discovering, accessing, and reusing research data across institutions and borders.

A key innovation in the plan is the concept of Data Labs within AI Factories — specialized environments that federate data from multiple sources and make it available for AI research in compliant, standardized formats. These labs address persistent challenges including data cleaning, enrichment, synthetic data generation, standardized formats, GDPR compliance, and interoperability between datasets from different institutions and countries.

The Commission also plans to improve access to and reuse of publicly funded research results, tackling the tension between open science principles and data protection requirements. For scientific AI applications, the quality and accessibility of training data is often the binding constraint — more important than model architecture or computational power. Europe’s strength in producing high-quality, curated scientific data represents a genuine competitive advantage that the action plan aims to leverage systematically. Researchers exploring related AI policy and strategy analyses will find complementary insights across the Libertify interactive library.

Building AI Talent and Addressing the Skills Gap

The human capital challenge is as significant as the computational one. The action plan identifies a critical shortage of researchers with combined AI expertise and domain science knowledge — the “hybrid profiles” needed to apply AI effectively in specific fields. The document calls for training the next generation of scientists to be AI-ready while simultaneously developing AI specialists with deep understanding of scientific domains.

The diversity gap is particularly acute. Women represent only 22% of AI professionals globally and account for just 13.8% of AI research paper authors. The action plan explicitly calls for strengthening inclusiveness and gender balance in AI research, though specific mechanisms beyond existing programs remain to be detailed. Research engineers, data stewards, and other supporting roles are also identified as critical but undervalued components of the AI research workforce.

Concrete measures include updating the European Competence Framework for Researchers (ResearchComp) to include AI proficiency, launching a new Self-Assessment Tool for AI skills, and funding Doctoral Networks on AI in science through the RAISE pilot. The Marie Skłodowska-Curie Actions and European Research Council will continue to play central roles in developing talent, while the EIT ecosystem’s approximately 800 AI startups with a total valuation close to EUR 20 billion provides a pathway for research-to-market transitions.

Trustworthy AI and Research Ethics Framework

The action plan’s emphasis on trustworthy AI distinguishes the European approach from competitors. Only 38% of Europeans trust scientific discoveries made with AI, according to Eurobarometer data. Meanwhile, 81% of researchers express concerns about AI models — specifically regarding ethics, accuracy, security, privacy, and lack of transparency. A further 63% cite the lack of guidelines as a factor hindering their adoption of AI tools.

The Commission’s response includes regularly updating Living Guidelines on responsible use of generative AI in research, commissioning the European Group on Ethics in Science and New Technologies (EGE) to deliver an opinion specifically on AI in science, and establishing an “ethics by design” approach as the standard for European AI research. The JRC is tasked with creating a Scientific AI Hub to monitor and evaluate AI models used in strategic research fields.

Research integrity risks from generative AI receive specific attention. The plan acknowledges concerns about fabricated citations, AI-generated plagiarism, and unverified text in scientific publications. These are not theoretical risks — they are already affecting scientific publishing and peer review processes. The Commission’s approach balances the need to enable AI adoption with safeguards against the degradation of scientific standards, framing trustworthy AI not as a constraint but as a competitive advantage that builds confidence in European research outputs.

Sector Applications: Advanced Materials and Biotechnology

The action plan highlights two strategic sectors where AI in science is expected to deliver transformative impact. The global advanced materials market is valued at approximately EUR 5 trillion, with the EU representing 25% of that market. AI adoption in materials science is growing at nearly 50% annually, and the combination of AI with robotic closed-loop synthesis systems could enable new material identification up to 1,000 times faster than conventional techniques.

In biotechnology, the EU sector represents approximately EUR 65 billion in turnover and 300,000 employees. The action plan uses AlphaFold as a compelling proof of concept: the AI system for protein structure prediction earned its creators the 2024 Nobel Prize in Chemistry and is now used by over 2 million researchers worldwide. Notably, this achievement depended on collaboration with the European Molecular Biology Laboratory (EMBL), demonstrating the value of European research infrastructure in enabling global AI breakthroughs.

Other sector examples mentioned include COSMIC-DANCE (machine learning for astronomy that discovered 70+ free-floating planets), F-IMAGE (AI for seismic signal analysis and earthquake research), AI-PREVENT (predictive and personalized healthcare), and BioMonitor4CAP (AI for biodiversity monitoring in agriculture). These examples illustrate the breadth of AI’s applicability across scientific domains and the Commission’s vision of AI as a horizontal enabler rather than a narrow specialty.

Startup Ecosystem and Innovation Pathways

The action plan addresses the research-to-market pipeline with specific provisions for AI startups and scale-ups in science. The EIT ecosystem currently includes approximately 800 AI startups with a combined valuation close to EUR 20 billion, providing a foundation for commercial exploitation of AI research outputs. The European Innovation Council (EIC) committed EUR 150 million to AI projects in 2024 and is positioned for an expanded role.

The EU AI Act’s treatment of scientific research is noteworthy: AI systems developed solely for scientific research and development are explicitly excluded from the Act’s scope. For research institutions and startups, this creates regulatory clarity — pure research activities are not subject to the compliance requirements that apply to commercial AI deployments. However, the Commission is analyzing the implications for research spin-offs, recognizing that the boundary between research and commercialization is often blurred.

AI Regulatory Sandboxes, mandatory in each EU member state under the AI Act, will provide controlled environments for testing AI innovations. The Commission also plans a pledging exercise for private sector investment in AI in science, seeking to complement public funding with corporate resources. For science-based AI startups, this combination of regulatory clarity, public funding, sandbox access, and the RAISE network creates a potentially attractive environment for development and scaling. Organizations tracking EU innovation and regulation developments will find this action plan a critical reference document.

Implementation Timeline and Next Steps

The action plan outlines a phased implementation with specific milestones. In Q4 2025, the RAISE pilot launches at the Copenhagen AI in Science Summit, accompanied by initial coordination structures, doctoral networks, thematic networks of excellence, and dedicated access to AI Gigafactories. Through 2026, the focus shifts to designing Data Labs, launching private sector pledging campaigns, and establishing monitoring frameworks for AI adoption metrics.

By 2027, the JRC Scientific AI Hub should be operational, providing systematic monitoring and evaluation of AI models in strategic research fields. The Commission will report on strategy implementation by end-2027, providing a formal assessment of progress and adjustments needed. The most ambitious targets are set for 2028 and beyond: fully building RAISE through partnership with member states and private sector, doubling AI investment levels, and developing RAISE under the new Multiannual Financial Framework.

International cooperation runs throughout the timeline. The Commission plans to shape AI in science norms through G7, G20, UNESCO, and OECD engagement while balancing openness with research security safeguards. This international dimension is critical: AI in science is inherently global, and Europe’s ability to attract international talent, participate in global collaborations, and influence international standards will significantly determine the strategy’s success.