Eval Injection in Langflow - CVE-2026-33873
Published: April 24, 2026
Vulnerability identifier: #VU127537
CSH Severity: Medium
CVSS v4.0: CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:U/U:Green
CVE-ID: CVE-2026-33873
CWE-ID: CWE-95
Exploitation vector: Remote access
Exploit availability:
No public exploit available
Vendor: Langflow
Affected software:
Langflow
Langflow
Detailed vulnerability description
The vulnerability allows a remote user to execute arbitrary code on the server.
The vulnerability exists due to improper neutralization of directives in dynamically evaluated code in the Agentic Assistant validation path when processing LLM-generated component code. A remote user can cause the model to return malicious component code to execute arbitrary code on the server.
The issue is reachable through the /assist and /assist/stream endpoints, and the streaming path reaches the sink only when the request is classified into the component-generation branch.
How to mitigate CVE-2026-33873
Install security update from vendor's website.