Vulnerability in LibreChat Affects Hosted AI Actions
CVE-2026-54044
7.7HIGH
What is CVE-2026-54044?
LibreChat, an advanced ChatGPT clone that integrates multiple AI providers, contains a flaw that allows a malicious collaborator with EDIT permissions to exploit the system. Before version 0.8.7, an attacker could manipulate a POST request to /api/agents/actions/:agent_id, which unintentionally exposes owner credentials when merging editor-controlled metadata with retained secrets. By redirecting actions to a server controlled by the attacker, sensitive information could be compromised. This vulnerability has been mitigated in version 0.8.7, emphasizing the importance of maintaining up-to-date software to prevent such risks.
Affected Version(s)
LibreChat < 0.8.7
