Regulatory Challenges Related to the Use of Artificial Intelligence for IAEA
Safeguards Verification

In the past decade, the International Atomic Energy Agency (IAEA) has recognized that artificial intelligence (AI) can provide important benefits for international nuclear safeguards through improved scope and performance, while reducing costs and manpower. But as AI is used in domains already governed by existing regulations, its uses for nuclear safeguards verifications will also have to be carefully conducted in line with regulatory requirements. The International Atomic Energy Agency (IAEA) has already been engaged in activities using AI for safeguards verification purposes and has expressed interest in further using this technology. A better understanding of the regulatory landscape for AI and how it may impact on nuclear safeguards verifications is imperative for determining whether these activities could potentially result in damage to the IAEA’s institutional interests and those of its Member States. The open-source country research conducted for this paper illustrates that in majority of jurisdictions, AI legislation and regulations have just recently started to be developed and that, in several instances, AI applications may conflict with the existing frameworks governing data protection and privacy, patent and copyright laws, as well as anti-discrimination policies. Therefore, the objective of this paper is threefold: 1) provide an overview of the AI legislative and regulatory landscape in several selected States; 2) identify certain potential legal risks and challenges pertaining to the use of AI for IAEA safeguards applications; 3) provide a series of recommendations for IAEA Member States to address these challenges.