AI Code Sandbox Escape Vulnerabilities
On 20 July, BleepingComputer reported sandbox escapes affecting Cursor, Codex, Gemini CLI, and Antigravity. The Hacker News reported that a DeepSeek Harness flaw let AI agents disable their own file sandbox without approval. It also reported that critical Cursor flaws allowed prompt injection to escape the sandbox and run commands; malicious .git configurations could make Claude, Codex, Cursor, and other AI agents execute attacker code; a Cohere AI Terrarium flaw enabled root code execution and container escape; and Google patched an Antigravity IDE flaw that enabled prompt-injection code execution.
BleepingComputer also reported that a critical vm2 sandbox bug allowed attackers to execute code on hosts. Another feed reported Snowflake Cortex AI escaping its sandbox and executing malware. OpenAI News described how it built a safe, effective sandbox to enable Codex on Windows. On 20 September, BleepingComputer reported that researchers had escaped the OpenAI Codex sandbox and run commands on the host. The reports describe tools designed to contain AI-generated code allowing that code to reach systems outside the sandbox.
Assuming the reports are factual, impact is 0%. The affected audience includes users and operators of Cursor, Codex, Gemini CLI, Antigravity, Claude, DeepSeek Harness, vm2, Snowflake Cortex AI, and Cohere AI Terrarium. The tracked event currently records no measurable impact or resulting change.
Watch for vendor advisories, patches, and technical disclosures that explain the affected versions and escape paths. Independent reproduction of the reported flaws would materially change the reliability assessment.
A flaw in DeepSeek Harness, DeepSeek's open-source tool for running AI coding agents on a developer's machine, let a sandboxed agent turn off its own sandbox with a single
Two flaws in Cursor, an AI code editor, could let a single, ordinary-looking prompt break out of the editor's safety sandbox and run any command on a developer's computer. There
A critical vulnerability in the popular Node.js sandboxing library vm2 allows escaping the sandbox and executing arbitrary code on the host system. [...]
Manifold Security has disclosed eight security flaws across seven command-line AI coding agents in which a repository's own Git configuration names a command that the agent runs
Learn how OpenAI built a secure sandbox for Codex on Windows, enabling safe, efficient coding agents with controlled file access and network restrictions.
A critical security vulnerability has been disclosed in a Python-based sandbox called Terrarium that could result in arbitrary code execution. The vulnerability, tracked as
Cybersecurity researchers have discovered a vulnerability in Google's agentic integrated development environment (IDE), Antigravity, that could be exploited to achieve code
NewsHive monitors these sources continuously. All signal titles above link to the original reporting.
Intelligence by NewsHive. Need help navigating what this means for your business? Contact GeekyBee →