New ENCFORGE Ransomware Targets AI Model Files in Langflow RCE Attack

What happened

Sysdig researchers linked a second attack on a Langflow server to JADEPUFFER, an AI-agent-driven ransomware operator identified earlier this month. This operator deployed a new strain of ransomware called ENCFORGE, which is written in compiled Go. ENCFORGE specifically targets files crucial to AI workflows, including model weights, vector indexes, and training datasets, encrypting them across the infected host’s filesystem.

The risk

ENCFORGE raises the stakes for AI infrastructure security. Encrypting core AI assets like model components and datasets can cripple operations that depend on them and force costly recovery efforts or payoffs. Since these resources are often large and complex, backup and restoration processes may be slow or incomplete, increasing downtime risk. The use of an AI-driven operator for intrusion and payload deployment adds automation and persistence capabilities, potentially accelerating attack speed and scale.

Why it matters

AI builders, operators, and businesses reliant on machine learning pipelines face higher risk of ransomware targeting the AI-specific parts of their infrastructure. This moves ransomware focus beyond generic file types to AI operational assets, which are harder to replace and more critical to business continuity. Teams must rethink security around model storage, version control, and dataset management to withstand escalating ransomware threats that specifically aim to disrupt AI workloads.

Who should pay attention

Security teams protecting AI infrastructure need to monitor for suspicious activity linked to AI-targeted ransomware. AI researchers and developers should safeguard access to training data and model files through stricter permissions, encryption, and segmented storage. Organizations deploying Langflow or similar open-source AI workflow tools must patch vulnerabilities and limit exposure to remote code execution exploits that can open doors for ransomware. Investors and operators should factor this risk into AI deployment strategies and vendor assessments.

What to watch next

Watch for further evolution of ENCFORGE and other ransomware strains designed to attack AI assets. Monitor vulnerability disclosures related to AI workflow tools like Langflow. Observe how defensive tools adapt to identify and block AI-targeted ransomware. The interplay between AI automation used by attackers and defensive response automation will shape AI infrastructure risk exposure going forward.

AI Quick Briefs Editorial Desk