Threat actors are actively exploiting two critical vulnerabilities affecting Langflow and Ruby on Rails, according to new findings released by VulnCheck. The security company reported that attackers are targeting the flaws to conduct credential probing, reconnaissance activities and command and control operations against vulnerable systems. The vulnerabilities include CVE 2026 0768, which carries a CVSS score of 9.8 and allows arbitrary Python code execution due to insufficient validation of user supplied input, and CVE 2026 66066, also known as KindaRails2Shell, with a CVSS score of 9.5. The second flaw can allow unauthenticated attackers to read arbitrary files from affected servers, exposing sensitive information including Rails secrets, database credentials, cloud storage credentials, API tokens and other confidential data that may ultimately lead to remote code execution. According to VulnCheck, attackers exploit CVE 2026 66066 by uploading specially crafted image files that abuse differences between Active Storage and libvips when processing image uploads from untrusted users.
VulnCheck reported that it recorded more than 50 detections within hours on August 30, 2026, with the total increasing to approximately 360 detections by Monday. Researchers observed attackers querying environment variables associated with Langflow, OpenAI API credentials, AWS access keys and AWS secret keys while also attempting to access cached secret files, SSH related information and shell history files. According to the company, the observed activity primarily originated from Russia and initially targeted monitoring systems located in the United Kingdom. The researchers stated that the activity indicates a combination of reconnaissance and credential harvesting designed to identify valuable secrets that could be used for further compromise. VulnCheck also noted that threat actors have exploited at least 12 vulnerabilities affecting AI related technologies since 2025, recording more than 15,000 successful exploitation attempts involving multiple Langflow vulnerabilities. Most vulnerable Langflow deployments identified by the company were located in the United States, Germany, Malaysia, Brazil and India.
Researchers also documented several attack scenarios involving previously disclosed Langflow vulnerabilities. In one observed case, attackers exploited CVE 2026 5027 to deploy a Python based credential harvesting tool, proxy agents and SimpleHelp remote access software on compromised systems. In another incident, threat actors abused CVE 2025 3248 to add compromised devices to an XMR cryptocurrency mining botnet. After establishing access, attackers disabled the Linux audit service to reduce forensic visibility before exploiting CVE 2026 0769 to deploy additional malware identified as .sysd. Researchers stated that the compromised systems were then used to scan for additional vulnerable hosts, suggesting an effort to expand cryptocurrency mining operations and identify further targets. These findings demonstrate the growing interest of threat actors in AI development platforms because they often provide access to cloud environments, sensitive credentials and interconnected infrastructure.
VulnCheck also detected active exploitation of CVE 2026 66066 against monitoring systems located in Singapore, Israel and the United Kingdom. According to the researchers, the observed activity originated from a single IP address in France and established command and control communications with infrastructure located in Israel. The company further noted that Active Storage enables libvips loaders that are considered unsafe when processing untrusted content, allowing specially crafted uploads to trigger malicious behavior. During testing, researchers found that although a patched Ruby on Rails 8.1.3.1 server successfully blocked the original libvips file read issue, another remote code execution mechanism involving variation key deserialization could still execute if provided with a valid signature. VulnCheck previously identified more than 7,100 internet exposed Ruby on Rails instances that remained vulnerable, highlighting the importance of applying security updates and reviewing exposed applications that process untrusted image uploads.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
