Executive Summary
Cato AI Labs disclosed DuneSlide, two Cursor flaws (CVE-2026-50548, CVE-2026-50549, CVSS 9.8) that let a single prompt escape the editor's sandbox and run commands without any click or approval.
What Happened
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 is no click to fall for and no approval box to ignore. Cato AI Labs found the pair and named them DuneSlide. They are tracked as CVE-2026-50548 and CVE-2026-50549, both rated 9.8 out of 10 (or 9.3
Affected Products
- Cursor
Vulnerability / Attack Technique
Prompt injection leading to sandbox escape
Technical Mechanism
As reported, crafted content processed by the editor's agent bypassed its sandbox and approval step; details are in the vendor advisory.
AI Security Category
LLM01:2025 Prompt Injection
Part of the Prompt Injection topic.
Relevant Frameworks
RealGround Analysis
Sandboxes and approval prompts are the main safety layer in AI IDEs, and both were bypassed. RealGround analysis: patch quickly and add an outer isolation layer for untrusted projects.
Recommended Controls
- Update Cursor to the fixed version
- Open untrusted projects in a VM or container
Related RealGround Service
Secure AI Agent Build · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/07/critical-cursor-flaws-could-let-prompt.html
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.