In short: JADEPUFFER uses ENCFORGE ransomware to specifically target AI model files following a second compromise of the same Langflow server.
Researchers at Sysdig have attributed ENCFORGE, a new ransomware written in Go, to a series of attacks on Langflow servers. The attacker JADEPUFFER deploys the malware strategically to encrypt model weights, vector indices, training data, and other AI infrastructure files.
Sysdig security researchers have attributed a second attack on the same Langflow server to the operator JADEPUFFER, which the company had already documented in early July. JADEPUFFER is known as an AI agent-driven attacker.
The same operator has now deployed ENCFORGE, a newly developed compiled ransomware based on Go. The malware is specialized in encrypting model weights, vector indices, training datasets, and further AI infrastructure files across the affected host’s file system.
For CISOs, this finding is critical because the compromise of Langflow servers via remote code execution (RCE) leads to direct access to production AI models. The targeting of model artifacts makes ENCFORGE a specialized threat to organizations operating AI systems in their infrastructure. Without deep knowledge of AI stack architecture, such ransomware deployment would be less likely; this indicates technical maturity on the part of the attackers.
Organizations should prioritize securing Langflow instances against RCE vulnerabilities, control access to model storage granularly, and implement dedicated ransomware detection for AI workloads.
Source: thehackernews.com · Published 21 July 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.3.