Securing Agentic AI by Ken Tannenbaum
AI agents do not just generate text. They execute actions, access systems, call tools, move data, and make decisions. The security model most organizations rely on was built for chatbots and APIs - not autonomous systems with permissions, memory, and workflow authority. The incidents are already here. Production databases have been deleted by coding agents ignoring explicit safeguards. Prompt-injection attacks have compromised trusted tooling ecosystems. Enterprise copilots have demonstrated zero-click data exfiltration vulnerabilities. Autonomous agents have been manipulated into financial loss through social engineering and tool abuse. Securing Agentic AI is the practical security guide for the agent era. Written for CISOs, security architects, security engineers, governance teams, and IT leadership, the book focuses on the real-world security, governance, and operational challenges emerging around agentic AI systems in 2026. It covers commercial tools, open-source frameworks, custom deployments, and multi-agent architectures already appearing inside enterprise environments. Inside, you will learn the modern threat landscape for agentic AI, including prompt injection, tool poisoning, memory manipulation, confused-deputy attacks, model exploitation, compromised MCP servers, A2A protocol abuse, excessive agency, and multi-agent cascading failures. The book unifies major frameworks including the OWASP LLM Top 10, OWASP Agentic Top 10, and MITRE ATLAS into a practical defensive taxonomy security teams can actually use. Dedicated sections cover Zero Trust architectures for AI agents, runtime enforcement boundaries, policy engines, identity and authorization frameworks, AI supply chain security, MCP vetting, data protection, observability, anomaly detection, and containment strategies for autonomous systems. Operational chapters focus on incident response for non-deterministic AI systems, forensic challenges, prompt-injection response playbooks, exfiltration investigations, compromised-agent recovery, and AI-specific red teaming using modern offensive tooling and evaluation frameworks. The book also provides detailed reference architectures for single-agent, multi-agent, customer-facing, and federated enterprise deployments, along with analysis of the rapidly evolving standards ecosystem surrounding agentic AI security. Included are deployment checklists, AI incident-response templates, prompt-injection test libraries, threat-model mappings, scoring worksheets, and extensive reference appendices. Thirteen chapters. Eight appendices. Built for organizations already deploying AI agents - or preparing for the reality that they soon will be.
| SKU | Unavailable |
| ISBN 13 | 9798950533204 |
| Title | Securing Agentic AI |
| Author | Ken Tannenbaum |
| Series | Ai For Everyone |
| Condition | Unavailable |
| Binding Type | Paperback |
| Publisher | Finnoybu Press |
| Year published | 2026-04-05 |
| Number of pages | 386 |
| Cover note | Book picture is for illustrative purposes only, actual binding, cover or edition may vary. |
| Note | Unavailable |