Learning from other Domains to Advance AI Evaluation and Testing: Testing in the Nuclear Industry

The growth in global energy demand accelerated in recent years (2023–2024) at rates double the average growth of the past decade (2010-2019). Projections suggest this trend will continue. Combined with the global goal of decarbonization, this positions nuclear energy as a reliable source with significant potential for growth. However, the benefits of the nuclear industry go beyond electricity generation and extend to healthcare, scientific and technological research and development, food security, and environmental protection among many other activities.
The nuclear industry is technologically mature and constantly evolving. The main associated risks are related to potential accidents and their consequences, including environmental contamination and radiological harms. Alongside the evolution of the nuclear industry, a consistent regulatory framework has been developed, evolving over time to keep up with new technologies and the operational experience of nuclear facilities. Different regulatory approaches exist in different jurisdictions. While some establish prescriptive regulations (i.e. defining how to do things), others use performance-based regulations (i.e. establishing a limit of risk and leaving the method of compliance to the designer). These approaches have in common that they establish standards aiming to ensure, with a high degree of certainty, that activities are conducted safely. In pursuit of this objective, testing provides an objective means of demonstrating compliance with specific requirements stated in regulatory or industry standards. However, testing is not only performed to ensure safety and regulatory compliance, it also supports and is an essential tool in maintaining the capability and availability of nuclear facilities to provide intended goods and services such as production of electricity, medical radioisotopes, neutron beams, etc.