Continual learning is a machine learning paradigm in which a system learns from new data over time while retaining useful prior knowledge. It aims to adapt to changing conditions without catastrophic forgetting of previously learned capabilities.