While Artificial Intelligence (AI) is having transformative effects in all areas of next-generation computing including cybersecurity, the idiosyncrasies of Critical National Infrastructure (CNI) raise profound questions over the role of AI in securing CNI.

Cyber-resilience of CNI requires accurate and informative detection and mitigation of cyberattacks combining data from multiple modalities into coherent models with explainable outputs, while respecting stringent latency and data sensitivity requirements as well as the broader (e.g., safety) context of diverse CNI sectors.

In this talk, we will explore how strategic use of AI approaches can address these challenges and how next-generation, converged IT/OT technology can be used to deploy such approaches into industrial networks to assure the resilience of diverse CNI environments including legacy.

Moreover, we will demonstrate how anomaly explanations can be provided through graph-based approaches to identify specific system components (e.g., sensors and actuators) that are being targeted during cyberattacks.