Critical Langflow Flaw Enables Remote Python Execution, Hackers Already Exploiting

Security researchers have confirmed that a high‑severity vulnerability in the open‑source AI workflow tool Langflow is being actively leveraged by threat actors. Identified as CVE-2026-0768, the flaw carries a CVSS rating of 9.8 and permits unauthenticated attackers to execute arbitrary Python code on any system running the affected version of the platform.

The issue stems from insufficient validation of user‑supplied input within Langflow's low‑code interface, which allows malicious payloads to be injected and run on the host server. Because the platform is often deployed in development and production environments to orchestrate large language model (LLM) pipelines, a successful exploit can give attackers control over critical data processing workflows.

Since the vulnerability was first disclosed by the security outlet Security Affairs, multiple intrusion attempts have been observed across a range of public and private deployments. Analysts note that the exploit does not require prior authentication, meaning that any internet‑exposed instance of Langflow could be compromised without the need for stolen credentials.

Langflow’s maintainers have responded by publishing a security advisory and releasing patches that tighten input handling and add optional hardening settings. The advisory urges users to upgrade to the latest release immediately, apply recommended configuration changes, and restrict network access to the service where feasible.

Industry observers say the incident underscores a broader risk associated with the rapid adoption of low‑code AI tools. While such platforms accelerate development, they often expose complex runtime environments to developers who may lack deep security expertise, creating attractive targets for cybercriminals.

Experts recommend that organizations using Langflow conduct a thorough inventory of all instances, verify that they are running patched versions, and monitor logs for anomalous Python execution patterns. Continued vigilance will be essential as attackers are likely to refine their techniques and seek out other similarly vulnerable AI‑focused applications.