The NEA Working Group on New Technologies (WGNT) hosted a workshop on March 25-26, 2026, to explore how artificial intelligence (AI) can be utilized by nuclear regulatory bodies for oversight and internal processes. The event brought together nuclear regulators and AI experts from 15 NEA member countries and international organizations to discuss and share case studies on AI tools already in use or under development, including applications for summarization, simulation enhancement, and information retrieval from regulatory documents.
Establishing an AI Regulatory Framework
A crucial takeaway from the workshop was the necessity for nuclear regulatory bodies to develop and implement a comprehensive AI framework. This framework should encompass clear procedures and detailed guidance for the responsible and effective integration of artificial intelligence into regulatory operations.
Importance of Well-Scoped AI Projects
Discussions highlighted that artificial intelligence initiatives yield the most effective and successful outcomes when the specific tasks and overall scope of the projects are precisely defined and focused from the outset. This ensures targeted development and application of AI tools.
Defining Clear Success Criteria for AI
Participants emphasized the imperative of establishing unambiguous and measurable success criteria for both individual AI tools and broader AI projects. This is essential for proper evaluation and to demonstrate their value and performance within the stringent regulatory context of nuclear safety.
Addressing AI Data Security and Privacy
A potential strategy to mitigate concerns regarding cybersecurity, data sovereignty, and data protection when deploying AI involves utilizing on-premise models and tools. This approach offers greater control over sensitive nuclear data and ensures compliance with security protocols.
Retaining Human Oversight in AI Applications
Despite the advancements in AI, the workshop underscored that human expertise remains absolutely indispensable. Human involvement is vital for critical decision-making processes and for intelligently interpreting and leveraging the outputs generated by artificial intelligence systems to ensure and improve nuclear safety.
