Identity and privilege abuse
An agent's credentials and permissions are used beyond what the user or the task allows: the agent acts as a confused deputy, widens its own access, or carries one user's rights into another user's request.
- ASI03:2026 Identity and Privilege Abuse
- LLM03:2026 Excessive Agency
- AML.T0053 AI Agent Tool Invocation
- AML.T0083 Credentials from AI Agent Configuration
- AML.T0098 AI Agent Tool Credential Harvesting
How it works
- 1Agents act with their own identity and tokens, which are often broader than any single task needs.
- 2A request the user could not make directly can succeed when the agent makes it on their behalf.
- 3Shared or long-lived credentials let access from one session spill into another.
What to watch for
- Actions outside the requesting user's own permissions
- Credentials used from new sessions or tools
- An agent asking for more scopes or roles
How to defend
- Give each agent the least privilege, scoped per task and per user.
- Pass the user's identity through and check authorisation on every action, not only at sign-in.
- Use short-lived credentials and keep them out of prompts.