What do you think?
Rate this book
Agentic AI Security: The Definitive Guide to Designing, Hardening, and Defending Autonomous LLM Agents Against Prompt Injection, Memory Poisoning, Tool Abuse, and Emerging Threats
Secure the future of autonomous AI. As large language model (AI) agents evolve from chatbots into goal-driven systems that reason, use tools, maintain long-term memory, and execute multi-step workflows, they introduce an entirely new attack surface. A single compromised prompt or poisoned memory can lead to data leaks, unauthorized actions, financial loss, or catastrophic failure.
Written by renowned AI security expert Andrew Ming, Agentic AI Security is the most comprehensive and practical handbook available for engineers, security architects, DevSecOps teams, red teamers, and responsible AI practitioners who build or defend autonomous LLM agents.
This step-by-step guide arms you with battle-tested frameworks, code patterns, and defensive architectures to protect production agentic systems today while preparing you for tomorrow’s threats.
Inside this definitive 10-chapter resource, you’ll agent-specific threat models with STRIDE-per-Agent, Attack Trees, and autonomous-system extensionsPreventing prompt injection, jailbreaking, and intent hijacking using schema-bound prompts, gated tool calling, and multi-stage validationSecuring short-term and long-term agent memory with integrity checks, encryption-at-rest, anomaly detection, and anti-poisoning filtersDefending against feedback-loop attacks, infinite recursion, self-modifying agents, and tool-abuse exploitsEmbedding real-time safety critics, policy engines, and intent alignment layers directly into the reasoning loopImplementing layered input/output sanitization, privilege minimization, human-in-the-loop guardrails, and runtime monitoringRed teaming autonomous agents with adversarial prompt libraries, automated fuzzing, and continuous threat simulationAchieving compliance with NIST AI RMF, OWASP Top 10 for LLM & GenAI, EU AI Act, and emerging global regulationsProduction-ready reference architectures for ReAct, AutoGPT-style, LangGraph, CrewAI, and custom agent frameworksFuture-proofing your agents against multi-agent swarm attacks, cross-agent contagion, and next-generation adversarial techniquesWhether you’re deploying customer-facing assistants, automated trading agents, internal enterprise workflows, or multi-agent research systems, this book gives you the exact playbooks used by leading AI labs and Fortune 500 security teams.
Don’t wait for the first breach. Master agentic AI security now and build autonomous systems that are intelligent, resilient, and unconditionally trustworthy.
Written by renowned AI security expert Andrew Ming, Agentic AI Security is the most comprehensive and practical handbook available for engineers, security architects, DevSecOps teams, red teamers, and responsible AI practitioners who build or defend autonomous LLM agents.
This step-by-step guide arms you with battle-tested frameworks, code patterns, and defensive architectures to protect production agentic systems today while preparing you for tomorrow’s threats.
Inside this definitive 10-chapter resource, you’ll agent-specific threat models with STRIDE-per-Agent, Attack Trees, and autonomous-system extensionsPreventing prompt injection, jailbreaking, and intent hijacking using schema-bound prompts, gated tool calling, and multi-stage validationSecuring short-term and long-term agent memory with integrity checks, encryption-at-rest, anomaly detection, and anti-poisoning filtersDefending against feedback-loop attacks, infinite recursion, self-modifying agents, and tool-abuse exploitsEmbedding real-time safety critics, policy engines, and intent alignment layers directly into the reasoning loopImplementing layered input/output sanitization, privilege minimization, human-in-the-loop guardrails, and runtime monitoringRed teaming autonomous agents with adversarial prompt libraries, automated fuzzing, and continuous threat simulationAchieving compliance with NIST AI RMF, OWASP Top 10 for LLM & GenAI, EU AI Act, and emerging global regulationsProduction-ready reference architectures for ReAct, AutoGPT-style, LangGraph, CrewAI, and custom agent frameworksFuture-proofing your agents against multi-agent swarm attacks, cross-agent contagion, and next-generation adversarial techniquesWhether you’re deploying customer-facing assistants, automated trading agents, internal enterprise workflows, or multi-agent research systems, this book gives you the exact playbooks used by leading AI labs and Fortune 500 security teams.
Don’t wait for the first breach. Master agentic AI security now and build autonomous systems that are intelligent, resilient, and unconditionally trustworthy.
147 pages, Kindle Edition
Published November 27, 2025
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
No one has reviewed this book yet.