CVE-2026-110930: OpenAI Codex Desktop for Windows Vulnerable to Code Execution via Malicious Library Paths

A high-severity vulnerability in OpenAI's Codex Desktop for Windows allowed attackers to execute arbitrary code through specially crafted Windows library paths, bypassing sign-in and project trust controls.

Christopher Ort
By Christopher Ort · 2 min read
Share
A Windows desktop computer displaying File Explorer with library folders open, photographed in a security operations center
On this page
  1. Overview
  2. Vulnerability Details
  3. Affected Versions
  4. Attack Requirements
  5. Discovery and Disclosure
  6. Mitigation
  7. Technical Classification
  8. Impact Assessment
  9. Sources

Overview

OpenAI has disclosed CVE-2026-110930, a high-severity vulnerability in Codex Desktop for Windows that could allow attackers to execute arbitrary code before users sign in or approve projects. The flaw, which carries a CVSS score of 8.0, stems from improper handling of quotation marks in project paths received from Windows File Explorer.

Vulnerability Details

The vulnerability affects how Codex Desktop processes file paths when users open projects from Windows libraries. According to OpenAI’s security advisory published on November 3, 2026, the application “did not correctly handle quotation marks in project paths from File Explorer.” This weakness allowed attackers to inject additional command-line arguments by embedding quotation marks in specially crafted paths.

When exploited, these malicious arguments could cause the application to execute attacker-controlled code with the permissions of the current Windows user, circumventing both the sign-in process and project trust approval mechanisms.

Affected Versions

The vulnerability impacts specific version ranges of Codex Desktop for Windows:

  • **Application versions**: 26.820.60940 through 26.1002.52244
  • **Microsoft Store package versions**: 26.820.7780.0 through 26.1002.7124.0

Attack Requirements

Successful exploitation requires several specific conditions:

  • Codex Desktop must be fully closed with no ChatGPT.exe process running
  • The user must open an attacker-controlled .library-ms file in File Explorer on Windows 11
  • The user must expand the malicious library in the navigation pane
  • The user must right-click an item under the library and select “Open project in ChatGPT” from the modern context menu

Discovery and Disclosure

The vulnerability was reported by Ikotas Labs, Inc. (@ikotas00) through the Zero Day Initiative. According to the advisory, the flaw was identified during Pwn2Own Ireland 2026, a competitive hacking event where security researchers demonstrate vulnerabilities in widely-used software.

OpenAI states there is no known abuse in the wild for this vulnerability.

Mitigation

OpenAI has released patched versions to address the vulnerability. Users should update Codex Desktop for Windows to:

  • **Application version**: 26.1007.21434 or later
  • **Microsoft Store package**: 26.1007.2314.0 or later

Technical Classification

The vulnerability is classified as CWE-88: Improper Neutralization of Argument Delimiters in a Command. This category covers flaws where applications fail to properly sanitize special characters that could alter the structure of commands or arguments passed to underlying systems.

Impact Assessment

With a CVSS score of 8.0, this vulnerability represents a significant security risk. The ability to execute code before authentication or project approval bypasses critical security controls designed to protect users from untrusted content. However, the attack complexity is elevated by the specific user interaction requirements and system state conditions needed for successful exploitation.

Sources

Christopher Ort

Christopher Ort

Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). He is the author of “The Style Penalty”, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.

Editor's picks

  1. CVE-2026-77604: Path Traversal Vulnerability in Code Graph RAG MCP Server Allows Arbitrary File Read
  2. CVE-2026-123271: ChatCLI OAuth Flaw Allows Token Theft via Malicious MCP Servers
  3. CVE-2026-123022: AgentScope MCP Workspace Allows Unauthenticated Remote Code Execution
  4. CVE-2026-97677: Path Traversal Flaw in IBM Langflow OSS 1.0.0–1.12.2 Allows File Read/Write
  5. Active Support DoS Flaw Lets Unauthenticated Attackers Crash Rails Servers via Action Cable