The adverse drug reaction (ADR) problem globally poses a significant threat to public health, by leading to substantial morbidity and mortality in patients as well as increased healthcare expenses. Current systems for pharmacovigilance rely primarily on spontaneous reporting methods that suffer from challenges related to the underreporting of ADRs, delays in signal detection, and a lack of capacity to adequately process the volume and complexity of expanding healthcare data. AI technologies (especially machine learning, deep learning, and natural language processing) have the potential to overcome these obstacles by providing automated means of analyzing very large and diverse datasets in the area of pharmacovigilance. This narrative review synthesizes the evidence from 20 published peer-reviewed studies from January 2015 through March 2026 that were identified by searching PubMed/MEDLINE, Scopus, Web of Science, and Google Scholar for studies assessing the use of AI methods in common pharmacovigilance functions such as automated ADR coding, signal detection, predicting individual patient risk for experiencing ADRs, and identifying drug-drug interactions. Anticipated major pharmacovigilance systems were reviewed including VigiFlow, VigiBase, the FDA's Adverse Event Reporting System, and EudraVigilance. Overall, based on the data from the studies reviewed herein, AI models were shown to have excellent predictive ability with AUC values reported between 0.77 and 0.96. However, challenges remain — including data quality limitations, issues with model interpretability, and the need to establish guidelines for regulatory approval of so-called "AI models" — yet AI has substantial potential to enhance pharmacovigilance and improve global drug safety monitoring.