Active Langflow and Ruby on Rails Exploits Open Door to Cloud Credential Theft and Remote Code Execution
Attackers are exploiting critical flaws in Langflow and Ruby on Rails, probing for cloud access keys in exposed AI workflow tools and abusing a Rails image-processing bug to leak secrets and run code on vulnerable servers.
Ongoing exploit campaigns against Langflow and Ruby on Rails are turning common development tools into entry points for cloud breaches and server compromises.
Security reporting says attackers are actively abusing a critical Langflow vulnerability to search for and extract OpenAI and Amazon Web Services credentials from exposed instances. Langflow is an open-source tool for building and orchestrating AI workflows; when left misconfigured or unpatched, it can reveal sensitive configuration fields that include live access keys.
In parallel, a separate flaw affecting Ruby on Rails applications that use the libvips image-processing library is being exploited to leak secrets and, in some cases, achieve remote code execution. The attacks focus on apps that accept untrusted file uploads, such as user images or documents. By crafting malicious image-like files, intruders can trigger vulnerable processing paths and gain deeper access.
Observed activity linked to the Rails and libvips issue has already moved beyond testing. Some attackers have established command-and-control channels on compromised servers, allowing them to maintain persistence, deploy additional tools or pivot into connected systems, including databases and internal services.
For teams running Langflow, stolen OpenAI or Amazon Web Services credentials can expose model interfaces, training data and other cloud resources to abuse. Experimental AI environments that were quickly deployed and poorly documented are particularly at risk of becoming blind spots in security monitoring.
For companies relying on Ruby on Rails, a successful remote code execution exploit in a file-upload pipeline can enable backdoors, data theft or manipulation of application logic. Services that handle sensitive documents, identity images or financial records are especially exposed.
The two campaigns highlight how attackers are targeting both AI orchestration layers and mature web frameworks to obtain high-value secrets and reliable footholds.
Key signals to watch are the release and adoption of patches or mitigation guidance for Langflow and affected Rails setups, any reported spikes in cloud abuse tied to compromised access keys, and signs that larger criminal groups begin scanning systematically for these vulnerabilities.
Sources
- OSINT