Back to blog
CVEAI & LLMRCEsupply chain

Langflow CVE-2026-9198: Unauthenticated RCE on the AI Platform Under Mass Exploitation

Published on 2026-08-056 min readCleanIssue

> TL;DR: CVE-2026-9198 (CVSS 9.8) is a code injection flaw in Langflow, a widely used open-source platform for building LLM-powered applications, that gives an unauthenticated attacker full remote code execution on default deployments. Fixed in July 2026 (version 1.10.1), it has already racked up 650 recorded exploitation attempts since July 6, from 244 unique IP addresses across 41 countries. CISA added it to its Known Exploited Vulnerabilities (KEV) catalog in early August.

Why this matters for SaaS vendors adopting AI

Langflow isn't an obscure product — it's one of the most popular tools for quickly prototyping and deploying applications that orchestrate LLMs (chatbots, agents, RAG pipelines). Many product teams use it to ship AI features without building orchestration from scratch. An unauthenticated RCE on a default deployment means any internet-exposed Langflow instance without hardened configuration is immediately compromised.

A history of repeated exploitation

This isn't an isolated incident. Langflow's security flaws have already been weaponized several times in recent months: a Monero cryptocurrency miner deployment, use by an autonomous AI agent to pivot toward other targets, and targeting by the EncForge ransomware, specifically designed to attack AI infrastructure. This pattern of repeated weaponization of the same platform reflects a broader trend: open-source AI tools, often deployed quickly with permissive default configurations, are becoming a prime target.

What to do

  • Update to Langflow 1.10.1 or later immediately if you're running this platform.
  • Never expose a Langflow deployment directly to the internet without upstream authentication (reverse proxy with SSO, VPN, or IP allowlisting).
  • Audit your logs for requests matching the known exploitation pattern, and check for unknown mining processes or web shells on affected instances.
  • The broader lesson for AI stacks

    AI application development platforms (Langflow, n8n, LiteLLM, and others we've already covered) often share the same risk profile: powerful, productivity-oriented tools deployed quickly by product teams eager to ship an AI feature, with authentication and access control as an afterthought. For any HR SaaS vendor integrating AI (resume parsing, candidate chatbot, payroll assistant), the security of the underlying orchestration platform deserves the same level of attention as the application itself.

    What CleanIssue checks for

    When auditing a stack that includes AI components, we systematically test the exposure and authentication of orchestration platforms (Langflow, n8n, and equivalents), not just the front-end application consuming them.

    Key takeaways

  • An AI orchestration platform exposed without authentication is a direct gateway into your infrastructure, not just a harmless prototyping tool.
  • Quickly deployed open-source AI tools are recurring targets — patch them with the same rigor as your critical applications.
  • Your AI stack's security should be judged end-to-end, from the application down to the underlying orchestration platform.
  • Building HR, payroll, or recruiting software? CleanIssue performs security audits for HR SaaS in real-world conditions, no source code access needed. For a first read of your exposure, start with an external review of your application.

    Need an external review of your HR SaaS?

    Share your product, stack, and client context. We will come back with the right review scope.

    Discuss your audit