AnythingLLM Desktop XSS in chat (CVE-2026-32626)
CVE-2026-32626
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, AnythingLLM Desktop contains a Streaming Phase XSS vu...
Overview
A critical security vulnerability, tracked as CVE-2026-32626, has been discovered in AnythingLLM Desktop. AnythingLLM is an application that allows users to chat with large language models using custom documents as context. This flaw enables a remote attacker to execute arbitrary code on a user’s computer simply by sending a specially crafted message in the chat interface.
Vulnerability Details
In versions 1.11.1 and earlier, the application contains a cross-site scripting (XSS) vulnerability in its chat rendering system. Specifically, the code that processes and displays images in chat messages fails to properly sanitize user input. This allows an attacker to embed malicious scripts within what appears to be a normal chat message or image reference.
Unlike other parts of the application that correctly sanitize output, the PromptReply component renders this tainted data directly into the webpage without cleansing. Because AnythingLLM Desktop is built on the Electron framework with insecure default settings, this script execution escapes the web view and gains the ability to run commands directly on the host operating system.
Impact
The impact of this vulnerability is severe (CVSS score: 9.6). With no required interaction beyond a user viewing a malicious chat message, an attacker can achieve full remote code execution. This could lead to:
- Complete compromise of the host system.
- Theft of sensitive data, including the documents loaded into AnythingLLM.
- Installation of malware or ransomware.
- Use of the system as a foothold for further network attacks.
This flaw is especially dangerous as it exploits the normal, intended use of the chat feature.
Remediation and Mitigation
The primary and immediate action is to update the software. Users of AnythingLLM Desktop must upgrade to a version newer than 1.11.1 immediately. The maintainers have released a patched version that properly sanitizes chat output and secures the Electron configuration.
If immediate updating is not possible, consider these temporary mitigations:
- Restrict Use: Avoid using the chat functionality with untrusted data sources or LLM connections until the patch is applied.
- Network Segregation: Run the application on a segregated or isolated system to limit potential lateral movement in case of exploitation.
For IT and security teams, this incident underscores the importance of securing Electron-based applications. Regularly review and harden the configuration of similar desktop apps in your environment. Staying informed on such vulnerabilities is crucial; you can find the latest cybersecurity news at security news. Furthermore, understanding how initial access flaws like this lead to major incidents is vital; review real-world examples in our breach reports.
Never miss a critical vulnerability
Get real-time security alerts delivered to your preferred platform.
Related Advisories
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, a SQL injection vulnerability in the built-in SQL Age...
SiYuan is an open-source personal knowledge management system. In versions 3.6.3 and below, Mermaid diagrams are rendered with securityLevel set to "loose", and the resulting SVG is injected into the ...
A Reflected Cross-Site Scripting (XSS) vulnerability exists in SourceCodester Sales and Inventory System 1.0. The vulnerability is located in the add_stock.php file via the "msg" parameter. The applic...
Notesnook is a note-taking app. Prior to version 3.3.11 on Web/Desktop and 3.3.17 on Android/iOS, a stored XSS in the Web Clipper rendering flow can be escalated to remote code execution in the deskto...