CISA KEV: Unauthenticated Code Injection in IBM Langflow (CVE-2026-9198) Actively Exploited

What Happened

CISA has added CVE-2026-9198, a code injection vulnerability in IBM's Langflow, to its Known Exploited Vulnerabilities (KEV) catalog as of August 4, 2026. According to IBM's advisory, the flaw allows unauthenticated attackers to achieve full remote code execution on default Langflow deployments. No CVSS score was published at the time of this writing, and details on the specific exploitation vector or scope of in-the-wild activity remain limited beyond CISA's KEV designation, which confirms active exploitation.

Why It Matters

Langflow is a visual builder used for developing LLM and AI agent workflows, and it has seen rapid adoption in AI/ML engineering environments. An unauthenticated RCE on a default deployment is a severe combination: it requires no credentials or user interaction, and any organization running Langflow with out-of-the-box configuration should assume exposure if the service is reachable. Given Langflow's role in orchestrating AI pipelines, successful exploitation could expose connected API keys, model credentials, data sources, or downstream systems the workflow integrates with, not just the host itself.

What Defenders Should Do Now

• Inventory all Langflow instances (internal and internet-facing) and determine version/patch status against IBM's advisory.
• Restrict network access to Langflow admin/API interfaces — do not expose default installations directly to the internet.
• Review authentication configuration; confirm the deployment is not running with default/unauthenticated access enabled.
• Hunt for anomalous process spawning, outbound connections, or unexpected code execution originating from the Langflow service account or container.
• Monitor for unusual requests to Langflow API endpoints, especially those associated with flow execution or code/component loading, and review logs retroactively for signs of prior compromise.
• Apply IBM's official fix or mitigation guidance as soon as it is validated in your environment.

Developing Intel

This is a same-day KEV addition and details are still emerging; CISA's listing confirms active exploitation, but public technical writeups and full remediation guidance may still be forthcoming. df00tech will update detection coverage as more information becomes available. See IBM's advisory for authoritative details: https://www.ibm.com/support/pages/node/7278927.