LIVE · cybersecurity feed
Live wire
OpenAI Announced $1B in Defensive Tools for Water UtilitiesAttackers Breached JetBrains Cadence via Unpatched TeamCity, Extracting AWS CredentialsCVE-2026-59346 · Critical VMware Workstation and Fusion Flaw Lets VM Admins Execute Host CodeCVE-2026-32475 · Elementor Pro WordPress Plugin Vulnerability Exploited to Hack SitesBroadcom Patches Critical VMware Workstation and Fusion VM-Escape VulnerabilitiesHackers Leak Millions of Airport Passenger Records After Ransom RefusalUsing a VM to Contain an AI AgentCVE-2026-73749 · HPE Patches Critical RCE Vulnerabilities in AOS-CXCVE-2026-14894 · Over 440,000 Exploit Attempts Target Super Forms and Elementor Pro RCE FlawsCisco searched for IOS XR bugs and found so many it rolled them into an update release
cve recordcritical

CVE-2026-50549

anysphere · cursor

Published
CVSS9.8
Severitycritical
WeaknessCWE-59
ExploitedNot in CISA KEV

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

Description

Cursor is a code editor built for programming with AI. Prior to 3.0, Cursor runs agent terminal commands in a sandbox by default. Before a Write, the agent canonicalizes the target path to confirm it stays inside the workspace, but when canonicalization fails it falls back to the original path and writes without approval. A malicious agent can create an in-workspace symlink that points outside the workspace and force canonicalization to fail — either because the target does not exist or because read permission is removed from the path — so the agent writes through the symlink to an arbitrary location without approval. A malicious agent could write arbitrary files outside the workspace under

References

← Back to the CVE Tracker

Our coverage of CVE-2026-50549

CVE-2026-12958high

Bug in top AI coding agents shows that Unix-era security headaches never really die

A vulnerability dubbed "GhostApproval" has been discovered in at least six popular AI coding assistants, allowing them to access files outside their designated workspaces and potentially execute remote code. The flaw exploits symbolic links, a long-standing security issue, to trick agents into writing malicious content, such as SSH keys, to sensitive system files. While some vendors have patched the issue and assigned CVEs, others have downplayed the risk or are yet to release fixes.