Executive brief
Langflow, a visual framework for building multi-agent and RAG applications, contains a critical vulnerability that allows an unauthenticated attacker to take full control of the server. By sending a specially crafted request to a specific system endpoint, an attacker can execute arbitrary commands with administrative (root) privileges. This could lead to the complete theft of sensitive AI models, data, and credentials, or the total disruption of the service.
Technical details
A remote code execution vulnerability exists in Langflow due to the inclusion of functionality from an untrusted control sphere (CWE-829). The flaw is located within the handling of the 'exec_globals' parameter provided to the '/validate' endpoint. An unauthenticated remote attacker can exploit this by submitting a malicious payload that is subsequently executed by the server. Successful exploitation allows for arbitrary code execution in the context of the root user. While version 1.4.2 is explicitly identified as affected, users are advised to update to version 1.9.0 or later to remediate the issue.
Affected products
- Langflow Langflow 1.4.2
Timeline
- 2025-07-18: disclosed: Vulnerability reported to vendor by ZDI
- 2026-01-09: advisory: Coordinated public release of ZDI advisory
- 2026-01-23: advisory: NVD publication date
- 2026-07-21: kev added: Added to CISA KEV catalog