ChatGPT-on-WeChat Agent Mode Vulnerability Exposes Users
A significant vulnerability, tracked as CVE-2026-6129, has been identified in zhayujie’s chatgpt-on-wechat CowAgent, affecting versions up to 2.0.4. According to the National Vulnerability Database (NVD), this flaw resides within an unknown function of the Agent Mode Service component and stems from missing authentication. This isn’t just a theoretical issue; the NVD has confirmed that an exploit is now public and actively being used, making this a live threat.
The lack of proper authentication means attackers can remotely manipulate the service without needing legitimate credentials. This is a classic blunder, often leading to unauthorized access or data exposure. The NVD noted a CVSS score of 7.3 (HIGH), underscoring the severity, with a vector indicating network-exploitable, low complexity, and no user interaction required. The project maintainers were reportedly informed of the problem via an issue report but have yet to respond, leaving users in the lurch.
Related ATT&CK Techniques
🛡️ Detection Rules
4 rules · 5 SIEM formats4 auto-generated detection rules for this incident, mapped to MITRE ATT&CK. Available in Sigma, Splunk SPL, Sentinel KQL, Elastic Lucene, and QRadar AQL.
Web Application Exploitation Attempt — CVE-2026-6129
Get this rule in your SIEM's native format — copy, paste, detect. No manual conversion.
4 Sigma rules mapped to the ATT&CK techniques from this breach — pick your SIEM and get a ready-to-paste query.
Get Detection Rules →Indicators of Compromise
| ID | Type | Indicator |
|---|---|---|
| CVE-2026-6129 | Auth Bypass | zhayujie chatgpt-on-wechat CowAgent up to version 2.0.4 |
| CVE-2026-6129 | Auth Bypass | Missing authentication in Agent Mode Service component |