Analysis · 22 October 2026

RAISE: European AI for science also needs to be governable.

At the European AI in Science Summit in Dublin on 16 October, the European Commission launched the RAISE (“Resource for AI Science in Europe”) pledging initiative. The signal is not the instant opening of a single platform: it structures contributions around the resources AI-enabled research needs.

Created on 22 October 2026 · Source reviewed: European Commission

1. What the Commission announced

The Commission presents RAISE as an initiative to strengthen AI capacity for science in Europe. It brings together actors able to contribute data, tools, skills, infrastructure and research practices. The announcement is neither a certification of tools nor a new obligation for businesses. It does make one fact clearer: industrial AI for science depends as much on governed resources as on the model selected.

2. What changes for Belgian and French organisations

For health, industrial, energy, chemical and food businesses, the opportunity is to bring research practices and production requirements closer together. An experiment becomes useful to the business only when datasets, access rights, model versions, results and human decisions remain traceable.

SMEs can use the emerging ecosystem to identify partners and reusable methods. Mid-market organisations, large companies and public bodies should prepare an exchange framework that protects trade secrets, personal data and intellectual property before sharing anything.

Underside’s view. AI for science should not create a governance exception. Whether it is a RAG system over publications, a laboratory-analysis agent or a flow connected to Odoo Enterprise, the decisive questions remain: which data is allowed, who validates it, where is it processed, which evidence is retained, and how can the service be exited?

3. An operational roadmap

separate public, collaborative and strictly confidential corpora;
document the provenance, licences and quality of data used by every model;
give agents querying a LIMS, ERP or Odoo dedicated identities and least privilege;
retain prompt, model, source and human-approval versions for sensitive outputs;
assess processing locations, subprocessors, support and reversibility before calling a chain sovereign.

4. Sovereignty is measured across the full chain

RAISE is a reminder that European scientific capacity is not just compute. For organisations in Belgium and France, a robust AI strategy connects infrastructure, data, models, security, skills and business accountability. European hosting or an open model can help, but neither replaces evidence of control across the whole chain.

Read the European Commission publication