EU to unveil RAISE, a CERN-style initiative for AI in science

Commission to outline EU plan for AI in science on 7 October

The European Commission will present a strategy next month to accelerate AI use across European research and, in return, use research to advance AI itself. The plan includes a concept and first actions for a new Resource for AI in Science (Raise) - a proposed "CERN for AI".

The announcement is expected on 7 October. For research leaders, this signals a shift from scattered pilots to a coordinated, EU-level approach.

What Raise could look like

• Shared compute and data infrastructure for research-grade AI work.
• Open model libraries and curated, license-clear datasets for science.
• Common standards for documentation, evaluation, and reproducibility.
• Training hubs and technical support for labs and research IT.
• Testbeds for safe deployment in sensitive domains.

The "CERN for AI" comparison suggests a collaborative, transnational facility focused on big scientific problems, not just generic tooling. For context on CERN's collaborative model, see CERN.

Why this matters for research teams

• Faster cycles: standardized resources reduce setup time and duplicated effort.
• Quality: shared benchmarks and protocols improve reproducibility and peer review.
• Access: pooled infrastructure may lower costs for institutions with limited capacity.
• Collaboration: common platforms make cross-border projects easier to run and govern.

What to do before 7 October

• Map your AI use cases across discovery, data management, analysis, and publishing. Rank by impact and feasibility.
• Audit data readiness: provenance, consent, licensing, retention, and security. Flag datasets that need cleaning or legal review.
• Inventory compute and storage needs for priority projects; estimate budgets with and without shared EU infrastructure.
• Define evaluation: target metrics, baselines, error budgets, and human-in-the-loop checkpoints.
• Prepare governance: model documentation, dataset cards, approval workflows, and incident response.
• Line up partnerships: identify labs, HPC centers, or industry groups you would join under a Raise framework.

Open questions to watch

• Scope: which scientific domains and which AI modalities will be prioritized first?
• Access model: who can use Raise resources, and how will time or credits be allocated?
• Funding: mix of EU, national, and institutional contributions; sustainability beyond initial pilots.
• Governance: oversight, ethical review, and safeguards for dual-use risks.
• IP and licensing: terms for datasets, models, and resulting publications or tools.
• Interoperability: alignment with open science policies, persistent identifiers, and existing research infrastructures.

How to position your team

Create a two-track plan: quick wins you can ship in 90 days, and 12-18 month programs that could plug into Raise. Document the dependencies where a shared EU resource would materially reduce cost or time, so you're ready to apply as soon as calls open.

If your team needs structured upskilling for implementation, explore focused options such as AI courses by job role or the latest AI courses.

Related context

The Commission has highlighted AI as a priority for research and innovation. For background on EU activity in this area, see the Commission's overview of AI in research and innovation.

Bottom line: the Raise concept points to a coordinated European backbone for AI in science. Use the weeks before 7 October to clarify your needs, governance, and collaboration strategy so you can move quickly when details land.