Scaling AI-Driven Knowledge Management and Knowledge Preservation

Effective knowledge management (KM) and knowledge preservation (KP) are critical for advancing the safety of reactor systems, radioactive waste management, and radiation protection research and training. To this end, the Commission supports, through the Euratom research and training programme, the development of a culture of knowledge sharing, reviews research, innovation, and training results, documents best practices, and monitors impacts. It has catalogued key knowledge outputs (deliverables, studies, reports, manuals, peer-reviewed papers) produced over the past 20+ years, forming a core Euratom-funded' knowledge base'. This database must be integrated with other internal and external sources to enable large-scale KM and KP and should also integrate related publicly available training material (masters and post doc level) from universities.

A crucial near-to medium-term task for Euratom is to establish a comprehensive Knowledge Management System. This system must move beyond today's multi-project basis to capture, store, share, and update the entire 20+ year knowledge base using modern tools such as semantic technologies, graph databases and enhanced visualisation techniques. This knowledge base could be enhanced and augmented through the use of appropriate AI models. Specifically, it should leverage AI, including proven AI-powered cognitive search, to improve querying and extract relevant information from diverse datasets, network maps, and social computing tools. Data security and regulatory compliance for sensitive information, as well as attention to intellectual property issues, are paramount. The system should also incorporate independent expert assessments covering basic to applied research, innovation, scientific and technological progress, and cross-cutting innovation. A user group could help to define the purpose, tooling and interface of knowledge management systems, that these will be used and useable for the longer term future.

In particular this action should establish an AI-driven knowledge system that indexes and preserves EC/Euratom R&I outputs and makes them LLM-searchable with source-grounded answers for policy and programme use. As a minimum, the AI functionality should cover: (i) answer questions by searching the database and always showing the source of the information; (ii) recognise important terms and connect them to standard nuclear concepts; (iii) organise information in a simple, structured map that shows where each piece of knowledge comes from; and (iv) include checks to make sure the answers are reliable, not misleading and free of bias.

This initiative will facilitate dissemination to both scientific and non-scientific/public audiences. It will strengthen priority alignment between the EU, Member States, and stakeholders, enable greater resource pooling at the EU level, support the integration of R&I results into policymaking with sectoral policymakers, and ultimately enable shared decision-making for strategic funding allocation to high EU-added-value projects.

The initiative involves scaling proven AI-powered cognitive platforms, knowledge bases, security measures, and expert networks for fission innovation. Coordination will occur primarily through the Euratom Research and Training Programme, leveraging internal/external international knowledge bases (e.g. EU/MS, OECD-NEA, IAEA).

The action should be implemented by a European consortium experienced in nuclear education and training, Euratom and national projects, taking stock from JRC having expertise and experience on preserving and disseminating European knowledge base. The consortium should create a sustainable, collaborative platform used and "owned" by the entire European nuclear research community, building a Community-Driven Ecosystem. The action requires specialized AI and industrial-grade IT platform expertise, leveraging diverse Expertise, enabling EU’s enhanced, secured and long-term KM and KP. The system should not duplicate existing international databases. Instead, it will be designed so that information from the EC Archive, EC R&D Information System (e.g. CORDIS, Funding and Tenders Participant Portal, JRC Science Hub, and the Publications Office of the European Union) and JRC relevant data can be connected and easily compared with other knowledge systems used by international partners. It should follow common European and international standards so that the results can be found and re-used both inside the Commission and in wider research networks.

The estimated 5-years’ CSA action should be composed of 3 phases: a) phase 1, a 6-month pilot project; b) phase 2, 6-month full implementation serving all community stakeholders; and c) phase 3 an effective and full implementation for the following four years, resulting in a sustainable, industrial-grade, AI-powered KM and KP platform for the long term. This will enhance expert peer-reviewed dissemination via tailored outputs for scientific, policy, and public audiences. To ensure appropriate coverage of research results and coordination, the actions should establish close links with Euratom's four co-funded partnerships: EURAD (radioactive waste), PIANOFORTE (radiation protection), EUROfusion (fusion), and CONNECT-NM (nuclear materials); European technology platforms and associations (e.g. SNETP, IGDTP, MEENAS, EERA-JPNM, SITEX), youth and professional societies (ENS-YGN, ENS), industry and regulators (nucleareurope, ENSREG), and the European Nuclear Education Network (ENEN).

Where appropriate, the Commission recommends that consortia use the services of the JRC. The JRC may participate in the preparation and submission of the proposal. The JRC would bear the operational costs for its own staff and research infrastructure operational costs. The JRC facilities and expertise are listed in General Annex H of this Work Programme.