Generative Artificial Intelligence and Large Language Models in Pharmacovigilance: Current Applications, Challenges and Future Perspectives.

Pharmacovigilance continuously monitors medicine safety by identifying adverse drug reactions and medication-related risks. The rapid growth of data from case reports, scientific literature, and electronic health records has significantly increased manual workloads. Consequently, Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) are emerging as pivotal tools to optimize information processing and interpretation. Current applications span data extraction, individual case safety report processing, case summarization, literature screening, signal and causality assessment, drug–drug interaction evaluation, and automated safety documentation. Evidence suggests these technologies enhance efficiency when handling vast datasets and repetitive tasks. However, critical challenges persist regarding hallucinated or inaccurate outputs, data biases, privacy concerns, lack of explainability, terminology errors, and insufficient real-world validation. Because pharmacovigilance drives safety-critical decisions, AI-generated outputs must never replace professional judgment. Human oversight, rigorous quality control, data protection, and regulatory framework integration remain vital. Future advancements in domain-specific language models, standardized evaluations, and human-supervised workflows will be essential to ensure the safe, prospective integration of GenAI into routine pharmacovigilance practice.