CISA Warns Langflow Flaw Is Actively Exploited to Hijack AI Workflows

· Original article ↗

Summary

CISA says CVE-2026-33017, a critical unauthenticated RCE in Langflow, is actively exploited. Endor Labs reports scanning began about 20 hours after disclosure; admins should upgrade to 1.9.0 or restrict the vulnerable endpoint.

Key points

  • CISA added CVE-2026-33017, a critical code injection flaw rated 9.3, to its Known Exploited Vulnerabilities list.
  • The flaw affects Langflow 1.8.1 and earlier and can enable unauthenticated remote code execution through a crafted HTTP request.
  • Endor Labs reports automated scanning began about 20 hours after disclosure, followed by Python-based exploitation and harvesting of .env and .db files.
  • CISA gave covered federal agencies until April 8 to patch or mitigate the flaw, or stop using the product.
  • Administrators are advised to upgrade to Langflow 1.9.0 or later, or disable or restrict the vulnerable endpoint.
  • Endor Labs also advises against exposing Langflow directly to the internet and recommends monitoring outbound traffic and rotating credentials if suspicious activity is detected.

Article Details

Vulnerability Types
  • Code injection
  • Unauthenticated remote code execution
Severity
Critical (9.3/10)
Affected Versions
  • Langflow 1.8.1 and earlier
Exploitation Status
active
Exploit Availability
claimed_or_private
Patch Status
available
Workarounds
  • Disable or restrict the vulnerable endpoint.
  • Do not expose Langflow directly to the internet.

MITRE ATT&CK

CVE

Products

Related Articles