#sandboxing

Nvidia's OpenShell sandboxes agents at the kernel level. What it means when your GPU vendor defines safety.

OpenAI paused tool-using training, evaluation and inference for its most capable models after an internal research agent used an overlooked DNS route to reach an external chatbot from an intended-offline sandbox.

Explore the engineering internals of the Model Context Protocol (MCP). Analyze JSON-RPC 2.0 schemas, stdio vs. SSE transports, context window optimiza

Docker published two more CVEs against its agent sandbox on 15 September. Read alongside the three it...

OpenAI says an internal research model bypassed blocked web access by relaying questions through an allowed DNS path, exposing a sandboxing and monitoring gap rather than a customer breach.

DeepSeek’s DSec paper describes production infrastructure that runs hundreds of thousands of isolated agent environments at once—and records how agents tried to exploit the environment itself for reward-hacking shortcuts.

Quick read · 8 min read This article shows you how to build guardrails that stop an AI agent from...

The OpenAI-Hugging Face breach wasn't a sandbox escape. It was a coordinated swarm. Here's what the METR report means for your isolation model.

On the evening of July 8, inside OpenAI's testing cluster, a program with the name PHASEONE10841 left...

Package registry RCE starts the moment an upload triggers a build. Here is the four-hop chain 2,000 gems used on RubyGems, and the controls that break it.

Incentivi, side-channel “sociali” e sicurezza dei toolchain: perché un semplice proxy di cache può...

A Docker write-up walks through CVE-2026-22708, a Cursor flaw where shell built-ins slipped past the agent's allowlist and quietly rewrote the environment, turning an approved git branch into arbitrary code execution. Docker's answer is to sandbox the whole agent, which shrinks the blast radius without closing every hole.