Langflow Exploitation Targets AI and Cloud Credentials; VulnCheck Adds CVE-2026-33309 to KEV
Active Exploitation · 3 sources · Updated 8d ago · First seen Oct 18, 2026
VulnCheck added Langflow vulnerability CVE-2026-33309 to its Known Exploited Vulnerabilities catalog after observing matching exploitation activity in its canary infrastructure. An actor operating from a Japan-based IP address targeted hosts in Israel with path traversal file-write probes matching CVE-2026-33309 and CVE-2026-5027, alongside attempts to exploit CVE-2025-3248, an unauthenticated code injection vulnerability. The source IP’s location does not establish the actor’s origin.
After gaining access through CVE-2025-3248, the actor searched environment variables and local directories for AI service and cloud infrastructure credentials, highlighting the risk of credential exposure beyond the Langflow application itself. The report identified 509 internet-exposed Langflow hosts and recommended removing public exposure and keeping the software patched. Defenders should also review affected deployments for exploitation and credential-access activity and rotate any potentially exposed secrets.
