// #sandbox-escape
4 articles
AI Coding Environments Join Pwn2Own Target List: LM Studio and OpenAI Codex Exploited via Sandbox Escapes
Pwn2Own Berlin 2026 introduced an AI products category and saw both LM Studio and OpenAI Codex exploited on the same day through sandbox escapes and environment variable injection. The results raise urgent questions about the security of AI development tools running inside enterprise environments with access to code repositories, credentials, and production pipelines.
Pwn2Own Berlin 2026 Day 1: Windows 11 Hacked Three Times, Edge Sandbox Escaped for $175K — 24 Zero-Days Demonstrated
The first day of Pwn2Own Berlin 2026 saw researchers demonstrate 24 previously unknown vulnerabilities across Windows 11, Microsoft Edge, VMware Workstation, and Oracle VirtualBox. Windows 11 was compromised three separate times by different teams, and a full Microsoft Edge sandbox escape earned a $175,000 award. No CVE IDs have been assigned yet as vendors begin the 90-day remediation process.
vm2 Node.js Sandbox Escape CVE-2026-26956 — 1.3 Million Weekly Downloads, PoC Published
A critical sandbox escape vulnerability in the vm2 Node.js sandboxing library allows a malicious script to break out of the sandbox and execute arbitrary code in the host Node.js process. CVE-2026-26956 affects all vm2 versions prior to 3.9.22 and is present in any application using vm2 to safely execute untrusted code — including serverless platforms, coding challenge sites, CI/CD systems, and plugin architectures. A PoC is publicly available.
Cohere Terrarium AI Sandbox Escape — CVSS 9.3 WebAssembly Flaw Allows Root Code Execution on Host
CVE-2026-5752 (CVSS 9.3) in Cohere Terrarium allows an attacker to escape the Pyodide WebAssembly sandbox via JavaScript prototype chain traversal, achieving root code execution on the host Node.js process. Organisations running AI code execution environments should patch immediately and network-isolate these workloads.