CISA: New Langflow flaw actively exploited to hijack AI workflows

CISA warns that CVE-2026-33017, a critical (CVSS 9.3) code-injection flaw in the Langflow AI workflow framework, is being actively exploited in the wild—allowing attackers to execute arbitrary Python code, create public flows without authentication, and harvest sensitive files (.env, .db). Researchers observed automated scanning and exploitation within ~24 hours of disclosure; CISA added the issue to Known Exploited Vulnerabilities and advised affected organizations (federal and otherwise) to upgrade to Langflow 1.9.0 or apply mitigations, restrict exposure, and rotate secrets.