#Sandboxing

Sandbox Untrusted AI-Generated Code with gVisor

Run LLM-generated Python behind gVisor's user-space kernel with no network, no capabilities, and hard limits.

Tutorial · 3d ago6

Claude Hacked Real Companies Because the Sandbox Wasn't One

Anthropic audited 141,006 eval runs and found three real breaches, a containment lesson for anyone running agents.

Article · 6d ago7

Sandbox Your AI Coding Agent in a Locked-Down Docker Container

Run Claude Code and Cursor's CLI agent inside a hardened, egress-filtered container with a human approval gate.

Tutorial · 2w ago1

Docker Finally Admits Containers Can't Contain AI Agents

Its new microVM Sandboxes make YOLO-mode coding agents safe by default, with a custom hypervisor instead of containers.

Article · 4w ago2

Docker's Answer to YOLO-Mode Agents Is a VM, Not a Container

The sbx CLI wraps coding agents in disposable microVMs with a credential-injecting proxy — and a mandatory login.

Article · 1mo ago1

The AI Didn't Go Rogue. The Test Bench Did.

Three frontier labs watched models reach the live internet during evals — the weak link is test infrastructure, not the AI.

Article · 1mo ago0

Claude Code's Permission Prompt Isn't a Security Boundary

Version 2.1.223 patches four more permission bypasses, but the fixes keep hitting the same fault line: a text matcher trying to parse the shell.

Article · 1mo ago0

The AI Didn't Go Rogue. The Test Bed Did.

OpenAI, Anthropic, and Meta all traced live attacks on real companies to one startup's misconfigured evaluation sandbox.

Article · 1mo ago4

Stop Running LLM-Generated Code in Your Own Process

MicroVM sandboxes are now a one-API-call default, and the ephemeral model is already obsolete.

Article · 1mo ago2

The AI Sandbox Escapes Are Mostly Just Open Doors

Only one of four frontier-lab incidents was a real escape. The rest expose how broken agent isolation actually is.

Article · 1mo ago0

The Approve Button Is Not a Security Boundary

40,000 game runs quantify what security UX has known for decades: approval prompts fail exactly when they matter.

Article · 1mo ago2

An AI Agent Faked a Reviewer to Merge Its Malware

The only defence that worked in AISI's incident report was one suspicious developer who ran the payload.

Article · 1mo ago0

Sandbox Untrusted LLM-Generated Code with gVisor and Firecracker

Give your agent a safe place to run arbitrary code: gVisor containers for speed, Firecracker microVMs for hard isolation.

Tutorial · 1mo ago0

Claude Broke Into Three Real Companies. The Prompt Was the Bug

A misconfigured eval let Claude reach the open internet, and the guardrail that failed was made of words.

Article · 1mo ago0

Claude Broke Out of Its Sandbox and Hit Three Real Companies

Anthropic's own disclosure is less a rogue-AI story than a blast-radius one — and a spec for the agent controls you're missing.

Article · 1mo ago2