Technical impact
IBM Langflow OSS versions 1.0.0 through 1.12.4 contain an improper control of code generation flaw that allows a remote authenticated attacker to execute arbitrary code. The flaw arises when user‑controlled data is passed to Langflow's code generation engine without adequate validation, enabling the attacker to inject malicious code. Successful exploitation can compromise the entire application, including data confidentiality and integrity, and provides the attacker full control over the host system.
Affected Systems
The vulnerability affects IBM's Langflow OSS, a machine‑learning workflow orchestration product, across all released versions from 1.0.0 up to and including 1.12.4. Any installation running these versions on any platform is potentially vulnerable. The official fix is available in version 1.12.5, which also includes previous fixes for CVE‑2026‑107366, CVE‑2026‑18894, and CVE‑2026‑77817.
Risk and Exploitability
The CVSS score of 9.9 indicates a critical severity with high exploit likelihood. Although EPSS data is not available, the lack of a KEV listing does not diminish the risk; the vulnerability permits remote code execution only after the attacker authenticates to the system, which is typical for high‑privilege accounts. Attackers would need valid user credentials, but the impact is still significant. Mitigation requires upgrading or applying configuration controls: hardening the MCP stdio interpreter and restricting admin‑only access to custom components. Additionally, revoking or rotating credentials that may have been exposed through exported flows is recommended.