Critical RCE flaw in Langflow already hit with 1,423 honeypot exploit attempts
Security firm Previdian identified active exploitation of a critical vulnerability in Langflow, an open-source platform that lets developers build AI applications and agentic workflows, chains of tasks an AI agent carries out on its own. The flaw, tracked as CVE-2026-105740 with a severity score of 9.9 out of 10, lets any logged-in user, even without admin rights, execute arbitrary commands on the server by adding a malicious MCP server, a protocol that connects an AI tool to external data sources, because the command field is passed straight to the operating system with no validation. Since September 19, Previdian has recorded 1,423 exploitation attempts against its own honeypot sensors, from 27 unique IP addresses across fifteen countries, with activity sharply increasing since October 11. Attackers typically log in using simple, guessable credentials and then install a persistent backdoor that re-executes every time the server list loads. For a business running internal AI tools on open-source platforms like this one, the case shows how risk often hides in configuration settings a normal code review would not catch.
Sources
This summary was written automatically with AI from the public sources listed above and checked by an independent AI reviewer before publication. For full details, read the original source.