CVE-2026-33017: critical vulnerability in langflow-ai langflow
Langflow has Unauthenticated Remote Code Execution via Public Flow Build Endpoint
Published · Updated
E-mail alert when it lands in KEV, gets a public PoC or its score goes up.
Patch now. It under exploitation confirmed by CISA and has a working public exploit.
Apply mitigations per vendor instructions, follow applicable BOD 22-01 guidance for cloud services, or discontinue use of the product if mitigations are unavailable.
Langflow allows anyone to run arbitrary code on the server through a public endpoint by uploading malicious workflow definitions without needing to log in. An attacker can execute any command they want on the machine running Langflow.
The POST /api/v1/build_public_tmp/{flow_id}/flow endpoint in Langflow versions before 1.9.0 accepts an optional data parameter containing attacker-controlled flow definitions with arbitrary Python code in node definitions, which are executed via exec() without sandboxing. Authentication is not required for this public endpoint, enabling unauthenticated remote code execution. The vulnerability was fixed in version 1.9.0 by removing acceptance of attacker-supplied flow data at this endpoint.
The full analysis of this CVE is available in Portuguese →
In the same product, most dangerous first.