CDM Inventor Unveils AI Defense Matrix for Generative AI Security

As enterprises adopt generative AI, security priorities have expanded beyond conventional networks and endpoints to include model weights, training data, prompts and agent identities. Cyber Defense Matrix (CDM) inventor Sounil Yu and SANS Institute security expert Lenny Zeltser developed the AI Defense Matrix to address traditional frameworks’ limited ability to represent AI-specific assets and define lines of responsibility.

The pair released the free framework in mid-May 2026, and iThome reported on it on May 26. The matrix uses a 6×8 design. Its horizontal axis aligns with the NIST CSF 2.0 functions Govern, Identify, Protect, Detect, Respond and Recover. Its vertical axis covers eight asset categories: AI workload platforms, orchestration tools, generated code, gateways, models, training data, runtime data and agent identities. Companies can use each cell to review controls and assign responsibilities.