Operational Risks and Reliability Challenges of AIClassifiers in Critical Infrastructure Protection

Artificial Intelligence classifiers are increasingly used in the protection of critical infrastructures,where security, resilience and service continuity are essential. These systems can analyse largevolumes of diverse data, detect anomalies, classify events and support faster operationaldecisions in critical sectors. The main objective of this narrative review is to analyse the role ofthese classifiers in critical infrastructure protection, with particular focus on reliability, resilienceand operational risk. The study examines the application of AI classifiers across several criticaldomains, including machine learning techniques for security classification, predictivemaintenance and anomaly detection. It further addresses governance requirements, operationalrisks and the implications of AI adoption for security auditing practices. Attention is given tochallenges that directly affect deployment viability, namely concept drift, adversarial attacks,false positives and negatives, lack of explainability, data dependency and integration constraintswith legacy systems. The role of human oversight, continuous monitoring, accountability andregulatory compliance is also considered, as these factors are increasingly central to theresponsible and effective use of AI in high-stakes environments.