HACKERS have begun exploiting a high-severity vulnerability in the Langflow low‑code AI platform to execute code remotely without authentication.
The flaw, tracked as CVE-2026-5027, carries a CVSS score of 8.8 and stems from improper sanitation of the filename parameter in API requests, allowing a path traversal that lets attackers write files to arbitrary locations on the host.
Once a malicious file is placed, the attacker can trigger its execution, achieving unauthenticated remote code execution on any exposed Langflow instance.
SecurityWeek notes that roughly seven thousand Langflow deployments reachable from the internet, mainly in North America, are at risk, and active exploitation attempts have already been observed.
The Hacker News highlights that no patch is currently listed for the issue, leaving administrators to rely on temporary mitigations such as blocking unauthenticated API access at the network layer and reviewing logs for suspicious file writes.
Defenders should immediately inventory public‑facing Langflow services, apply network‑level restrictions to limit API exposure, and monitor for unexpected file creation or process spawning as indicators of compromise.