External risk intelligence

DeepChat Stored XSS to RCE Vulnerability

CVE advisorySeverity: CRITICAL (CVSS 9.6)

CVE-2025-66222

DeepChat is a desktop-based smart assistant application built on Electron. Its attack surface is local to the user's machine, as it is a client-side application rather than a public-facing web service or network appliance. The vulnerability requires local interaction within the application context, making internet-facing exploitation via common network exposure mechanisms highly unlikely.

Code Injection

Thinkinai Deepchat

0.5.0 and earlier

Halo Surface Signal: 1 out of 5 — much less likely to be public-facing.

External exposure likelihood

Horizon Alert

Summary of the vulnerability and why it matters

This advisory highlights a critical vulnerability in DeepChat, an AI-powered smart assistant application. The issue involves the way the application renders certain diagrams, which could allow an attacker to execute malicious code on a user's system, potentially leading to broader system compromise. The main concern is confirming if this application is in use and, if so, understanding the extent of its deployment.

  • Stored XSS in diagram rendering allows code execution.
  • Escalated to RCE by exploiting Electron IPC.
  • Confirm relevance and exposure within the organization.

Attack Path

How an attacker could exploit the issue

An attacker could target users of DeepChat by tricking them into interacting with a specially crafted message containing malicious code. This code exploits a flaw in how the application renders diagrams, allowing JavaScript to run within the app. If successful, this could then be used to execute arbitrary code on the user's computer.

  • Requires no prior access to the system.
  • Triggered by viewing a malicious diagram.
  • Allows arbitrary code execution.

Live Threat

Current exploitation, exposure, and threat context

This vulnerability could allow an attacker to execute arbitrary JavaScript within the application's context when a specially crafted Mermaid diagram is rendered. When supported, this could be escalated to remote code execution through the Electron IPC bridge by registering and starting a malicious MCP server.

  • Arbitrary JavaScript execution in application context.
  • Malicious diagram rendering via user interaction.
  • Potential for remote code execution.

Operational Fix

Recommended remediation, mitigation, and detection steps

For DeepChat, application owners and platform teams are likely responsible for addressing this stored XSS and RCE vulnerability. The first practical step is to identify all instances of DeepChat, confirm their business criticality and exposure, and then assign an accountable owner before planning remediation activities.

  • App owners should manage remediation.
  • Verify Electron IPC bridge usage.
  • Plan maintenance for upgrades.

Supplementary metadata

Validate whether this threat affects your internet-facing exposure.

Halo Threat Intelligence helps prioritize remediation with Halo Surface Signal and H/A/L/O context. Start exposure validation with a free external attack surface trial.

Frequently asked questions

What is DeepChat?

DeepChat is an AI-powered smart assistant application built on the Electron framework. It is typically used as a desktop client to help users interact with artificial intelligence models, often featuring capabilities like diagram rendering to visualize complex information or data flows.

What does CVE-2025-66222 mean?

This vulnerability involves Stored Cross-Site Scripting (CWE-79) and Improper Control of Generation of Code (CWE-94). An attacker can inject malicious scripts into a Mermaid diagram. Because DeepChat runs on Electron, this script can interact with internal system bridges to execute unauthorized commands, escalating to Remote Code Execution.

How is this vulnerability triggered?

The issue is triggered when a user views a specially crafted Mermaid diagram within the application. Simply having the software installed is not enough to trigger the bug; the attacker must successfully entice a user to interact with or render the malicious content containing the exploit payload.

Is my organization at risk from this?

According to Halo Surface Signal, this risk is very unlikely for network-based threats. Because DeepChat is a local desktop application, it does not function as a public-facing web server. Exploitation generally requires local user interaction, making it different from vulnerabilities that allow remote attackers to scan and target systems over the internet.

What steps should I take if I use DeepChat?

Begin by inventorying all systems where DeepChat is installed to understand your footprint. Once identified, assign ownership to coordinate testing and verify that you are running the latest, patched version of the software. Prioritize these systems based on their business use and the potential impact of local code execution.

References