#9
BleepingComputer
general
July 20, 2026 at 21:14 UTC
Cursor, Codex, Gemini CLI, Antigravity hit by sandbox escapes
By Ax Sharma
AI Summary
Researchers discovered sandbox escape vulnerabilities in four AI coding tools — Cursor, Codex (OpenAI), Gemini CLI, and Antigravity — by inducing AI agents to write files that trusted host tools subsequently execute, resulting in host-level code execution. Multiple CVEs were issued and patches released, but the attack class exposes a systemic trust boundary problem in agentic coding environments.
Relevance score: 82.0/100
Sponsored
Protect Your Business
Expert cybersecurity solutions to safeguard your organization from evolving threats.
Get Protected →