Attackers exploit path traversal flaw in AI platform Langflow

Attackers have begun actively targeting CVE-2026-5027, a high-severity path traversal vulnerability in the popular AI development platform Langflow, using it to write arbitrary files on exposed servers. This flaw, which carries a CVSS score of 8.6, allows unauthorized remote actors to bypass intended directory restrictions and place malicious files on vulnerable systems. Security researchers first identified the exploit in the wild earlier this year, noting that the attack chain typically begins with a simple HTTP request that manipulates file paths. Once a threat actor gains write access, they can deploy web shells, modify application configurations, or plant backdoors for persistent access. The vulnerability resides in Langflow’s file upload and processing endpoints, which fail to properly sanitize user-supplied paths. Organizations running Langflow versions prior to the latest patch are at immediate risk. The platform’s maintainers have released a security update that resolves the flaw by enforcing strict input validation and directory whitelisting. Any team using Langflow for building or deploying AI workflows should verify they are on the most recent version and check server logs for signs of unauthorized file writes or suspicious directory traversal attempts. Given the active exploitation, delaying this update could expose sensitive data and infrastructure to compromise.