Least privilege for agents
Also called: principle of least privilege, PoLP
Granting an agent only the tools, data access, credentials, and autonomy its current task needs, and nothing beyond that.
The agentic version of a foundational security principle. In practice: scope the agent's identity to specific resources, connect only the servers the task needs, use read-only credentials where writes aren't required, keep sensitive data out of context unless it's needed, and require approval for the few actions that are irreversible.
Why it matters
It is the one defense that works no matter how the model was compromised, because it limits consequences rather than trying to prevent causes. Every documented toxic flow can be broken at some privilege boundary. When time is limited, tightening what the agent can do beats adding another detector.
Exposure map
Live counts of instances showing this pattern will appear here once the exposure map is collecting data.