Excessive agency
Also called: excessive permissions, over-privileged agent
Giving an agent more capability, permission, or autonomy than its task requires, enlarging the damage when it misbehaves.
Named in the OWASP Top 10 for LLM Applications. It has three flavors: too many tools (functions the task never needs), too many permissions (a tool whose credentials can do far more than the tool's purpose), and too much autonomy (high-impact actions with no human confirmation). Each one widens the blast radius of a prompt injection.
Why it matters
Most agent-security advice reduces to constraining agency: fewer tools, tighter scopes, approval gates on irreversible actions. It is the most reliable lever because it limits impact regardless of how the model was tricked. You usually cannot prevent the injection; you can decide what it is able to accomplish.
Exposure map
Live counts of instances showing this pattern will appear here once the exposure map is collecting data.