Memory is stored context that lets an AI system retain information beyond a single interaction. In practice, it can include user preferences, task history, prior outputs, or other state that the system reuses later. That makes an AI assistant more useful for long-running work, but it also turns one prompt into part of a larger workflow.
In cyber security, memory matters because anything retained may be abused or leaked. An attacker can try to poison memory with false instructions, trigger the reuse of stale data, or cause sensitive details to persist longer than intended. Defenders reduce risk with strict retention rules, separation of user sessions, access controls, encryption, logging, and clear ways to review or delete stored context. Good memory design keeps continuity without turning the system into a repository for secrets or attacker-controlled state.