AIBound

Home / Glossary / Excessive agency

Glossary

What is excessive agency?

Excessive agency is the risk that an AI system can take harmful actions because it was given more tools, more access or more freedom than its task needs. It is LLM06:2025 in the OWASP Top 10 for LLM Applications. OWASP names three root causes: excessive functionality, excessive permissions and excessive autonomy.

By the AIBound team · Updated 17 November 2026

Key takeaways

  • Excessive agency means an AI system can do more than its task needs.
  • OWASP lists it as LLM06:2025 and names three causes: too much functionality, too many permissions and too much autonomy.
  • The fix is fewer tools, narrower access and a person's approval before high-impact actions.

Why excessive agency matters

An AI model makes mistakes. It can misread a request, invent a fact or follow an instruction hidden in a document. When the model can only write text, a mistake is a wrong answer. When it can call tools, a mistake becomes an action.

Excessive agency decides how bad that action can be. An agent that can only read one folder can do little harm. An agent that can delete files, send email and change production systems, with nobody approving each step, can do a great deal.

How excessive agency happens

OWASP describes three root causes. A single agent often has more than one.

  1. 1 Excessive functionality: the agent has tools or extensions the task does not need, such as a delete function on a read-only job.
  2. 2 Excessive permissions: a tool connects to other systems with more access rights than the task requires.
  3. 3 Excessive autonomy: the agent takes high-impact actions without a check or a person's approval.

Excessive agency vs prompt injection

Prompt injection is a way to trick an AI into following an attacker's instructions. Excessive agency is what makes the trick costly. One is the attack and the other sets the size of the damage.

Excessive agency vs prompt injection
Question Excessive agency Prompt injection
What is it? An AI with more tools, access or freedom than it needs Input that makes an AI follow an attacker's instructions
Where does it come from? How the agent was set up Text the AI reads, such as a prompt, a web page or a file
How is it reduced? Fewer tools, narrower access and approval steps Treating outside content as untrusted and checking actions

Sources

Sources checked 17 November 2026

Examples

Excessive agency examples

FAQ

Frequently asked questions

What is excessive agency?

Excessive agency is the risk that an AI system can take harmful actions because it has more tools, access or freedom than its task needs. OWASP lists it as LLM06:2025 in the Top 10 for LLM Applications. The harm usually starts with a model error or a manipulated input.

What causes excessive agency?

OWASP names three root causes. Excessive functionality means the agent has tools the task does not need. Excessive permissions means those tools hold more access to other systems than the task requires. Excessive autonomy means the agent takes high-impact actions without a check or a person's approval.

How do you prevent excessive agency?

OWASP's guidance is to keep extensions and their functions to the minimum, give each one the least permissions it needs, run actions in the user's own context and require a person's approval for high-impact actions. In short: fewer tools, narrower access and an approval step where the stakes are high.

How does AIBound help limit excessive agency?

AgentTracer maps what each agent can reach and surfaces agents with more reach than their job needs. AgentHarness then enforces your policy inside the agent, so each action is allowed, asked about or denied in real time. The ask step puts a person's approval in front of a risky action.

Stop the risky AI your employees already run.

AIBound blocks the high-risk AI on every employee device, in real time. The 14-day trial needs no proof of concept.

Book a Demo