RC RANDOM CHAOS

OpenShell Sandboxes AI Agents With Kernel-Level Policy Enforcement

· via Hacker News

Original source

OpenShell – A safe, private runtime for fleets of autonomous AI agents

Hacker News →

NVIDIA introduced OpenShell, a runtime environment designed to safely execute autonomous AI agents. The system grants agents the ability to read files, install packages, call APIs, and use credentials while preventing unauthorized access through strict access controls.

OpenShell enforces permissions using kernel-level isolation combined with policy validation. Each agent runs in a sandbox where kernel controls restrict file access and system call permissions, and network traffic undergoes policy checks before leaving the sandbox. Credentials remain hidden from agents and are injected only into requests destined for approved endpoints. Before policy changes take effect, the system uses formal verification to flag potentially dangerous new access, such as credential-enabled connections to unapproved hosts, requiring human approval for such changes.

The project targets Linux, macOS on Apple Silicon, and Windows with WSL 2 in experimental status. It is distributed under the Apache License 2.0 and includes components for sandbox management, policy configuration, credential handling through providers, and Kubernetes integration.

Read the full article

Continue reading at Hacker News →

This is an AI-generated summary. Read the original for the full story.