A background agent is software that continues performing tasks outside an active user session. It may check connected data or run scheduled actions at intervals, then notify the user when something needs attention. Unlike a chatbot that responds to prompts, it can work between interactions using permissions granted in advance.
In cyber security, this creates an ongoing trust boundary: the agent may access private information, interpret external content, and sometimes initiate actions. Excessive permissions, misleading input, or unclear approval rules can lead to unwanted exposure or activity. Defenses include granting only the access needed, requiring human approval for consequential actions, reviewing activity logs, and revoking access when it is no longer needed. Background work is not necessarily continuous monitoring; its frequency and capabilities depend on the system.