VulnerabilityBleeping Computer
7.5 — HIGH
Cursor, Codex, Gemini CLI, Antigravity hit by sandbox escapes
Researchers escaped the sandboxes in Cursor, Codex, Gemini CLI and Antigravity by having the AI agent write files that trusted host tools later run. Multiple CVEs, patches, and Google downgrading two Antigravity findings. [...]
🤖 AI BriefingAuto-generated threat analysis
🔍Threat Overview
Four widely used AI coding agents, including Cursor, OpenAI's Codex, Google's Gemini CLI and Antigravity, were breached by sandbox escapes, allowing malicious code execution on host systems without direct attack on the sandbox.
⚙️Technical Details
CVEs
CVE-2026-48124
Affected Systems
CursorCodex CLIGemini CLIAntigravity
Attack Vectors
workspace-controlled .claude hook configtrusted local tool running on agent-written fileprivileged local daemon
💥Impact Assessment
Severity: high
Who Is at Risk
users of affected AI coding agentsorganizations with multiple users of these agentsSeverity: high
🛡️Recommended Actions
1Implement strict access controls for workspace-controlled files and hooks.
2Regularly review and update trusted local tool configurations to prevent agent-written file execution.
3Monitor system logs for suspicious activity related to privileged local daemons.
📦Affected Products
CursorCodex CLIGemini CLIAntigravity
🔐NVD Verified DataVERIFIED
Weaknesses
CWE-94CWE-829
Read the full article
This is a curated summary. The complete article is available at Bleeping Computer.