A multi-agent system contains several software agents that perceive an environment and take actions with some independence. The agents may cooperate, compete, divide work, or operate separately while affecting shared resources. Their interactions can produce capabilities and risks that do not appear in a single-agent test.
Designers must account for communication, identity, permissions, incentives, conflict resolution, and collective behavior. Security evaluation should observe the population and its shared infrastructure over time rather than assuming that controls proven for one isolated agent will remain effective when many agents interact.
