Safely run AI-generated code in Cloud Run sandboxes

Google Cloud Run sandboxes provide a secure, isolated environment to run untrusted code and agent workloads, eliminating the need for complex sandboxing infrastructure. This feature is crucial for developers who need to execute AI-generated code or user-submitted code without compromising their host application or data. To use Cloud Run sandboxes, simply enable the sandbox launcher via gcloud or YAML configuration. This allows for near-instant execution of untrusted code, with average latency of 500ms. Cloud Run sandboxes support use cases such as LLM code interpreters, headless browsers, and user-submitted code execution.