Critical Langflow Flaws Allow Unauthenticated Remote Code Execution
Original source
Critical Langflow Flaws Could Let Unauthenticated Attackers Run Code on Servers
The Hacker News →Two critical vulnerabilities in Langflow, an open-source platform for building AI agents and workflows, allow unauthenticated attackers to bypass authentication or execute arbitrary code. IBM disclosed the flaws in a bulletin published October 28-29. All versions 1.0.0 through 1.12.8 are vulnerable; version 1.13.0 (released October 22) provides the fix.
The first flaw enables remote code execution through Langflow’s Model Context Protocol (MCP), which allows flows to be offered as tools to other AI applications. A publicly available Metasploit module exploits the vulnerability (CVE-2026-93680), connecting to the MCP endpoint without credentials, establishing a session, and enumerating available tools—though the module does not execute code on the target server. The module’s creator lists versions 1.5.0-1.11.4 as affected, a narrower range than IBM’s assessment. Exploitation was demonstrated with automatic login enabled.
The second flaw represents incomplete remediation of an auto-login security control. Langflow’s LANGFLOW_AUTO_LOGIN setting is enabled by default in version 1.12.8, automatically authenticating users as superusers without requiring credentials. This setting should only be used in development environments. Version 1.13.0 disables automatic login by default.
IBM simultaneously published 29 additional CVE records on October 29, with fixes in versions 1.12.4 and 1.12.5. One of those, CVE-2026-104339, also allows unauthenticated code execution and is rated 9.8 in severity. Neither of the two critical flaws appears in CISA’s Known Exploited Vulnerabilities catalog as of October 30, though six earlier Langflow vulnerabilities are listed there.
Read the full article
Continue reading at The Hacker News →This is an AI-generated summary. Read the original for the full story.