The AI lifecycle is the full sequence an AI system goes through: data collection and preparation, training, testing, deployment, and ongoing monitoring or retraining. Each stage can introduce security and privacy risks if controls are weak.
In cyber security, the lifecycle matters because attackers often target the data path, not just the model itself. Poisoned training data, exposed prompts, leaked outputs, or overly broad access to connected records can all change how an AI system behaves or what it reveals. Defenders use the lifecycle as a framework for governance: classify data, enforce least privilege, verify provenance, monitor model and data drift, and review logs for abuse. Treating AI as a lifecycle problem helps organizations secure both the model and the information it uses.