Summary
This document is a guide that outlines 100 controls for enhancing the security of Agentic AI systems. It details various governance and risk management strategies aimed at mitigating specific risks associated with AI implementations. Each control is presented with its implementation steps and the risks it aims to mitigate. For instance, the guide discusses the importance of executive accountability, data classification, and prompt registry management, among others. It emphasizes the need for robust measures such as output handling guardrails, model selection policies, and human oversight training to ensure safe and secure AI operations. The document also covers technical aspects like tool integrity monitoring, security analytics for AI signals, and adversarial red teaming, providing a comprehensive framework for organizations to adopt in order to safeguard their AI systems against potential threats and vulnerabilities. Overall, the guide serves as a resource for organizations looking to implement effective security practices in their AI deployments.
