How do multi-agent systems multiply governance risk?

Answer

Multi-agent systems create compound risk through:

1. Delegation chains — Agent A delegates to Agent B, which calls Agent C. Who authorized the final action?
2. Emergent behavior — agents interacting in ways not predicted by individual testing.
3. Cascading failures — one agent's error propagated and amplified across the system.
4. Attribution — when something goes wrong, determining which agent in the chain caused the failure.
5. Permission inheritance — ensuring delegated agents don't accumulate broader permissions than the originating agent. Governance must track the full execution graph, not just individual agents.