New ransomware targets AI model weights and can't even collect the ransom

A sophisticated ransomware variant called ENCFORGE is specifically targeting AI model weights and training data—assets that cannot be quickly restored from backups—with recovery costs estimated at $75,000-$500,000 per model, representing a material business risk that traditional cybersecurity frameworks fail to address. The attacker exploits unpatched vulnerabilities (CVE-2025-3248) to gain persistence, demonstrating adaptive attack capabilities that can pivot strategies in minutes when initial approaches fail, indicating a shift toward lower-cost initial compromise followed by opportunistic lateral movement. This threat demands IT organizations align security investments with quantifiable business impact on AI assets rather than treating it as a purely technical cybersecurity issue, requiring new backup and recovery strategies specifically designed for machine learning infrastructure.