The first workshop of NEA’s AI Platform for Nuclear Research and Education (AIxpertise), held online from October 15 to 16, 2025, contributed to the further development of a new collaborative project dedicated to harnessing artificial intelligence (AI) for the benefit of nuclear energy professionals.
With AIxpertise, NEA aims to strengthen collaboration between stakeholders in the nuclear sector from industry, safety, research, and academia, and equip professionals with AI knowledge and equipment to advance research and train the next generation of nuclear professionals.
The workshop brought together 45 organizations from 17 countries, including safety agencies, research institutes, academia, industry, and big tech companies. The program included presentations and discussions on the overall project structure, NEA IT and Data Bank Services, the legal framework for the AIxpertise project, and the draft program of work across three focus areas: data, AI algorithm benchmarking, and education and training.
Participants discussed ongoing efforts and explored ways to strengthen and coordinate them through collaboration within the AIxpertise program.
data
The Jožef Stefan Institute (IJS) in Slovenia announced the pulse and depletion experiments carried out at the IJS TRIGA Mark II research reactor and plans for the development of a new versatile European reactor VERONICA for neutron irradiation and nuclear research. IJS has offered to share TRIGA reactor data with AIxpertise members.
Norway’s Institute of Energy Technology (IFE) has published a review of the Halden Reactor Project (HRP) legacy database, an important resource covering over 60 years of experimental work essential to advances in research and safety assessment.
A team at the Nuclear Research Data System (NRDS) at the Idaho National Laboratory (INL) has demonstrated search augmentation generation (RAG) that has the potential to significantly improve the accessibility and interpretability of HRP data content.
Oak Ridge National Laboratory (ORNL) in the United States has demonstrated optimized machine learning-driven exploitation of existing datasets to improve modeling and simulation (M&S) and validate M&S in areas of sparse validation data, such as the High Analysis Low Enriched Uranium (HALEU) application case, which is an important candidate for future small modular reactor deployments.
Participants discussed the scope of the AIxpertise project towards fostering data that meets the FAIR principles (searchability, accessibility, interoperability, and reusability) as the foundation for AI&ML approaches. They expressed strong support for AI-driven tools efforts to promote accessibility and interpretability of data, including HRP data. There was also strong interest in operators’ proposals to share research reactor data and nuclear power plant time series data with AIxpertise members protected by the AIxpertise legal framework.
Benchmarking AI algorithms
The benefits of international collaboration in benchmarking activities were outlined in a presentation on the results of the NEA Nuclear Science Committee Task Force on Artificial Intelligence and Machine Learning for Scientific Computing in Nuclear Engineering. North Carolina State University (NCSU) has expressed support for the development of international benchmarks within AIxpertise.
Presentations by Imperial College in the UK on Fundamental AI for Fluid, Solid, Particle and Radiation Modeling Methodologies and INL in the US on Machine Learning for Nuclear Fuel and Materials Modeling demonstrated new opportunities to improve modeling capabilities and accuracy with AI applications.
Microsoft has provided a glimpse into the future of Agentic AI, offering the potential to improve efficiency and speed up the nuclear permitting process by fully integrating AI agents into human-driven workflows.
Participants discussed the scope of the proposed AIxpertise project, which aims to benchmark the output of AI algorithms and ensure their transparency, reliability, and readiness for regulatory review.
Practical training and best practices
The Tokyo Institute of Science discussed the need for the development and comprehensive validation of knowledge management systems based on assistive AI agents based on RAG technology. North Carolina State University in the United States and Politecnico di Milano in Italy presented existing educational opportunities and expressed support for developing and delivering training to AIxpertise project participants. Participants discussed the benefits of different training formats to foster the development and continuous learning of nuclear talent.
The workshop concluded with a roadmap to launch the AIxpertise project in early 2026. The next step will be to refine AIxpertise’s work program based on further feedback from potential project participants and review of legal agreements.
Interested organizations are invited to participate in the development of AIxpertise by contacting AIxpertise@oecd-nea.org. The next AIxpertise workshop is scheduled for January 21-22, 2026.
For more information, please visit the Aixpertise page.
