Friday, 16 October 2026
1 agent hack today▼ 7 vs yesterday (8)

Attackers target critical Langflow flaw in 1,423 exploit attempts

Security firm Previdian says honeypots recorded more than a thousand attempts to exploit a maximum-severity bug in the open-source AI workflow platform.

By The Agentic Times ·

Attackers target critical Langflow flaw in 1,423 exploit attempts
· Image: thehackernews.com

Threat actors are actively exploiting a critical flaw in Langflow, the open-source platform used to build AI applications and agentic workflows, according to an alert from security firm Previdian reported on 15 October 2026.

The vulnerability is tracked as CVE-2026-105740 and carries a CVSS score of 9.9 out of 10, placing it at the top of the severity scale. It is described as an authenticated remote code execution flaw, meaning an attacker who can authenticate to an exposed Langflow instance may be able to run their own code on the underlying machine.

Previdian said its honeypots — decoy systems set up to attract and record attacks — logged 1,423 exploit attempts against the flaw. That figure indicates opportunistic scanning and exploitation at scale rather than isolated probing, although Previdian's alert as reported does not break down how many of those attempts succeeded or who was behind them.

Langflow is maintained by DataStax, which was acquired by IBM in February 2025, according to The Hacker News.

The platform matters to the agent security picture because of what it does. Langflow is used by developers to wire together the components of AI applications and agentic workflows, which typically means it holds the credentials, model API keys and data connections those workflows depend on. A system that orchestrates agents is, by design, a system with reach into other systems. Code execution on such a host is therefore a different proposition from code execution on an isolated web server.

The "authenticated" qualifier is important but should not be read as reassurance. Authenticated bugs still matter where instances use default or weak credentials, where credentials have leaked, or where an organisation has exposed a development tool to the public internet on the assumption that a login page is sufficient protection. The volume of attempts recorded by Previdian suggests attackers consider the barrier worth testing.

Organisations running Langflow should check whether their deployments are reachable from the internet and apply vendor guidance for CVE-2026-105740. At the time of Previdian's alert, no details of affected version ranges or patch availability were included in the reporting seen by The Agentic Times. Administrators should treat any internet-facing Langflow instance as a priority for review while that information is confirmed.

Sources

  1. thehackernews.comhttps://thehackernews.com/2026/10/langflows-critical-authenticated-rce.html