A critical unpatched vulnerability in Langflow, an open-source low-code platform for building AI applications, is being actively exploited in the wild. Tracked as CVE-2026-5027, the flaw is a path traversal issue that enables an attacker to write files to arbitrary locations on exposed servers. VulnCheck and BleepingComputer both confirmed the exploitation activity.
The vulnerability carries a CVSS score of 8.8, reflecting its high severity. According to reports, exploitation does not require authentication, widening the pool of at-risk systems. Any Langflow instance exposed to the internet without mitigations is potentially compromised, though exact numbers of affected deployments remain unknown.
Technical analysis reveals the flaw resides in the 'POST /' endpoint, where insufficient input validation allows directory traversal sequences. By sending crafted requests, an attacker can upload malicious files—such as web shells—to arbitrary directories. Once placed, these files can be executed remotely, granting full control over the host. No specific indicators of compromise have been publicly shared yet.
No official patch has been released for CVE-2026-5027. As a temporary workaround, administrators are urged to restrict network access to Langflow instances, placing them behind firewalls or VPNs. Blocking external access to the API endpoint and monitoring for unusual file uploads are also recommended until a fix is available.
Attribution for the attacks has not been established. The incident underscores the growing risks in the AI development toolchain, where low-code platforms frequently expand the attack surface. Given the active exploitation and lack of a patch, organizations using Langflow should treat this as an urgent security event.