How to meet it
OWASP AISVS v1.01
Infrastructure, Configuration & Deployment Security · AI Workload Sandboxing & Validation
Run models in a sandbox
Verify that AI models execute in isolated sandboxes.
OWASP AISVS v1.01, requirement 4.1.1
In plain English
Run models in an isolated environment with only the access they need, so a compromised model process cannot reach the rest of your systems.
What good looks like
- Model servers run in containers or VMs with restricted network access.
- The process has no credentials beyond what serving needs.
- File system access is limited to the model files.
How teams usually do it
- Containers with read-only file systems, dropped privileges and network policies.
- Separate VMs or gVisor-style sandboxes for untrusted models.
Evidence an auditor would accept
- Container and network policy configuration.
Common mistakes
- Running the model server with broad cloud credentials attached.
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.