How to meet it
OWASP AISVS v1.01
Orchestration & Agentic Security · Component Isolation and Tool Authorization
Allow-list what the agent installs or calls
Verify that external resources named in model output are verified against an approved allow-list or registry before the agent installs or invokes them.
OWASP AISVS v1.01, requirement 9.3.7
In plain English
If model output names something to install or call, such as a package, URL or plugin, check it against an approved list before the agent acts.
What good looks like
- Packages, URLs and plugins are checked against an allow-list or registry.
- Unknown items are blocked.
- Blocks are logged.
How teams usually do it
- A package proxy that allows approved packages only, and URL allow-lists in the tool gateway.
Evidence an auditor would accept
- Allow-lists and gateway configuration.
Common mistakes
- Letting an agent install any package the model suggests, including made-up names an attacker has registered.
Not sure which requirements apply to your system? Take the assessment.
Requirement text from the OWASP AI Security Verification Standard, licensed under CC BY-SA 4.0. This guidance is shared under the same licence.