Definition (as of 2026-10-07)
Agent Privilege Surplus is the reach an agent exercises above the reach it declared, measured per right (files, network, execution, secrets, persistence, approval) on the Privilege Reach scale from 0 to 3. Where a right is not declared, the tool’s task scope (level 1) is the yardstick:
surplus = max(0, observed − scope), with scope = the declared level, or 1 where the right is not declared.
Privilege Reach is the scale itself: each tool is read three ways, Declared (what the author says), Code (what static analysis finds) and Lab (what one lab run showed). The reach level of an entry is the maximum of its known code values.
The German umbrella term is Schatten-Administration (definition).
What it is not
- Not shadow AI. There, “shadow” means unknown or not inventoried. A surplus can come from a known, inventoried agent.
- Not a privilege gap. That term is used by several vendors for other measures.
- Not granted versus used. Permission analytics measure which granted rights are used. The surplus measures used against declared. Whether a right is needed cannot be determined reproducibly; that stays a human judgement.
- Not “excessive agency” in the OWASP sense, which describes a vulnerability class, not a measured difference.
Cite as
Ilja Schlak. “Agent Privilege Surplus”. Definition as of 2026-10-07. Agentic Security Field Guide. https://readbeforerun.ai/terms/privilege-surplus/