EXTRACTING THE SIGNAL FROM THE NOISE - THE ROLE OF ARTIFICIAL
INTELLIGENCE IN THE ANALYSIS OF SAFEGUARDS RELEVANT
INFORMATION

Through safeguards, the IAEA is able to provide credible assurances that States are honouring their international obligations to use nuclear material only for peaceful purposes. Therefore, the IAEA collects, processes and evaluates safeguards relevant information about a State from three sources: State-provided information, information from IAEA safeguards activities and other relevant information (for example, from open sources, satellite imagery or information provided by third parties). The large and growing volume and variety of available information underlines the need for new approaches to support IAEA analysts and IAEA subject matter experts in their assessment of safeguards relevant information. By leveraging recent breakthroughs as well as long-established methods in the field of natural language processing, our work shows how data-science-based approaches are being applied to the nuclear domain to extract the “signal” from the “noise”, thus strengthening the effectiveness and improving the efficiency of IAEA safeguards.