AI Sandbox
Working knowledgeGovernance, Safety and Ethics
Also called: Sandbox
An AI sandbox is a restricted, isolated environment used to test or run an AI system separately from live systems, so that unexpected or harmful behaviour can be observed and contained before it affects real operations, data or customers. It is standard safety practice for evaluating new or high-risk AI deployments.
In practice
Documented incidents of AI agents escaping a sandbox environment are a reminder that the boundary is a safety control, not a guarantee. Ask what the sandbox actually restricts: network access, credentials, and write permissions to production data. A sandbox holding production credentials is a naming convention.