DeepChat versions before 1.0.4-beta.1 contain a critical improper input validation vulnerability (CWE-20) that allows arbitrary protocol execution bypass resulting in remote code execution (RCE). The vulnerability arises because the patch for CVE-2025-55733 failed to sanitize native Electron pop-up window handlers. An attacker or compromised AI endpoint can craft a Markdown link with target="_blank" that triggers a native window interception in tabPresenter.ts, forwarding the malicious URL directly to shell.openExternal(url) without passing the isValidExternalUrl security check. This bypass enables execution of arbitrary protocols, compromising confidentiality, integrity, and availability. The issue is resolved in version 1.0.4-beta.1.

Successful exploitation allows remote attackers or compromised AI endpoints to execute arbitrary protocols on the victim's system via Electron's shell.openExternal function. This leads to a complete compromise of confidentiality, integrity, and availability of the affected system running DeepChat versions prior to 1.0.4-beta.1. The vulnerability has a CVSS 3.1 base score of 9.6 (critical), indicating high severity with network attack vector, low attack complexity, no privileges required, user interaction needed, and scope change.

Upgrade DeepChat to version 1.0.4-beta.1 or later, where this vulnerability is fixed. Since the vendor advisory confirms the fix in this version, applying this update fully mitigates the risk. There is no indication of partial or temporary fixes, and no alternative mitigations are provided. Patch status is confirmed by the vendor's versioning information.