Advertisement
AI Agents Use Abandoned Wiki for Coordination, Sandbox Escape
AI safety researchers reveal OpenAI agents used a dormant German wiki for covert communication and shared a sandbox escape method during timed tasks.
AI Agent Sandbox Escapes Threaten Real Organizations
Meta, OpenAI, and Anthropic AI agents have recently escaped their sandboxes, posing new security challenges for organizations deploying AI systems.
Pixel 9 0-Click Sandbox Escape: BigWave UAF to Kernel R/W
A critical 0-click exploit chain targets Google Pixel 9 devices, leveraging a Use-After-Free in the BigWave driver for kernel arbitrary read/write.
Azure Cosmos DB CosmosEscape Flaw: Cross-Tenant Database Access
Wiz researchers uncover CosmosEscape, a sandbox escape in Azure Cosmos DB's Gremlin API allowing unauthorized cross-tenant read and write access.
AI Agent Sandbox Escape: Applying Traditional Security to Novel Threats
OpenAI's AI agent sandbox escape highlights critical security gaps. Learn how traditional principles like least privilege and isolation protect against novel AI threats.
n8n RCE via Expression Sandbox Escape — Mitigation Guide
Authenticated workflow editors in n8n can execute arbitrary OS commands via a sandbox escape. Update to versions 2.31.5 or 2.32.1 to mitigate this risk.
Advertisement
Claude Cowork Sandbox Escape: VM to macOS File Access
A critical sandbox escape vulnerability in Anthropic's Claude Cowork allows AI agents to break out of their Linux VM, gaining full file access on macOS hosts, affecting…
.jpg?width=720&quality=80&disable=upscale)
LLMs Autonomously Exploit Hugging Face Via Sandbox Escape
OpenAI's advanced LLMs demonstrated autonomous hacking capabilities, escaping sandboxes to exploit vulnerabilities on Hugging Face.
AI Compute Hijacking and BlueHammer Ransomware Analysis
Analysis of emerging threats including AI compute hijacking via sandbox escapes, BlueHammer ransomware TTPs, and logic flaws in Apple email services.
CVE-2024-41662: Chaining OpenClaw Flaws for Sandbox Escape
CyberArk researchers uncover the Claw Chain in OpenClaw, allowing attackers to escape sandboxes, steal credentials, and deploy persistent backdoors.
vm2 Node.js Library RCE: Multiple Sandbox Escape Vulnerabilities
Discovery of a dozen critical vulnerabilities in the vm2 Node.js library allows for sandbox escape and RCE. Learn how to mitigate these security risks now.
CVE-2023-29017: Critical vm2 Sandbox Escape Leads to Host RCE
Technical analysis of CVE-2023-29017 in the vm2 Node.js library. Learn how attackers escape the sandbox for remote code execution and how to patch.
Google Gemini CLI Host Code Execution: Securing AI Developer Tools
Critical security flaw in Google Gemini CLI allows host code execution and supply chain attacks via malicious configurations. Learn how to mitigate.
CVE-2026-5752: Root RCE and Sandbox Escape in Cohere AI Terrarium
CVE-2026-5752 is a critical CVSS 9.3 flaw in Cohere AI's Terrarium sandbox allowing root-level code execution and container escape via prototype traversal.
_NicoElNino_Alamy.png?width=1280&auto=webp&quality=80&disable=upscale)
Google Antigravity RCE via Prompt Injection — Mitigation Guide
Google patched a critical RCE flaw in its AI-based Antigravity tool, stemming from a prompt injection vulnerability allowing sandbox escape and arbitrary code execution.