Blog

AI Agent Glossary

Sandboxing

Running an agent's actions in an isolated environment so mistakes cannot reach production systems.

Sandboxing gives the agent a disposable container, VM or browser profile with scoped credentials. Coding agents run tests inside it; browsing agents use isolated profiles; cloud agents get short-lived tokens. It limits blast radius from both model errors and prompt injection, and is a prerequisite for raising autonomy level.

Source: OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments — Xie et al., 2024

See the numbers behind this term: AI Agent Statistics 2026.