LLM Scalability Risk for Agentic-AI and Model Supply Chain Security Published in the Journal of Computer Information Systems (2026)
-
Keywords: AI Governance, Agentic-AI, LSRI, Model Supply Chain Security, Cybersecurity, Dual-use AI.
Overview: As Large Language Models (LLMs) transition from static chatbots to autonomous Agentic-AI, they introduce unprecedented scalability risks and security vulnerabilities. This paper provides a comprehensive analysis of the GenAI-driven cybersecurity landscape, examining both offensive (AI-driven malware, social engineering) and defensive (automated threat detection) perspectives.
Key Contributions:
-
LLM Scalability Risk Index (LSRI): A new parametric framework designed to stress-test and quantify operational risks when deploying LLMs in security-critical environments.
-
Model Supply Chain Framework: A proposed methodology for establishing a verifiable "root of trust" across the entire model lifecycle to ensure integrity and safety.
-
Defense Synthesis: An analysis of modern defensive strategies from platforms like Google Play Protect and Microsoft Security Copilot, culminating in a governance roadmap for secure, large-scale AI deployment.
