An architecture where multiple AI agents collaborate or compete to accomplish tasks. Multi-agent systems create complex liability and attribution challenges because harm may result from emergent interactions rather than any single agent's action.
See: Agentic AI; Orchestration