liveBook · Manning

What you will learn in AI Governance

Secure, privacy-preserving, ethical systems

by Engin Bozdag and Stefano Bennati · Manning Early Access Program (MEAP)

Chapter by chapter

Chapter 1: Governing Generative

The chapter introduction explains why GenAI’s rapid, flexible adoption makes practical governance urgent.

1.1 Why Governance, Risk and Compliance (GRC) for GenAI Matters Now? · 1.2 GRC for AI: More Than a Compliance Checklist · 1.3 A Mental Model for GenAI GRC · 1.4 Challenges in Practice: Illustrative Scenarios · 1.5 Tools and Practices You’ll Need · 1.6 Motivation and What’s Ahead · 1.7 Summary

What you learn in chapter 1 →

Chapter 2: Adoption Models for GenAI

The introduction establishes why an organization must identify how it adopts GenAI before deciding how to govern it.

2.1 Three postures on the adoption spectrum · 2.2 SaaS or Application Consumers · 2.3 API Integrators · 2.3.1 Organization-Controlled Zone: Enhanced Customization and Governance · 2.3.2 Vendor-Controlled Zone: Stability and Reliability · 2.3.3 Governance and Risk Management Opportunities · 2.3.4 Achieving Effective Governance for an API Integrator · 2.4 Model Hosters · 2.4.1 Architecture Overview: What’s Inside the Controlled Zone · 2.4.2 Governance and Risk Management Opportunities · 2.4.3 Achieving Effective Governance for a Model Hoster · 2.5 Risk dimensions across deployment postures · 2.6 Agentic AI · 2.6.1 What Does “Agentic” Mean? · 2.6.2 Common properties of Agentic AI · 2.6.3 Why Agentic AI is still experimental · 2.6.4 Governance Challenges · 2.7 MediAssist: A case study in AI governance · 2.8 Summary

What you learn in chapter 2 →

Chapter 3: The Six Level GenAI Governance Framework

This introduction uses three incidents to show why GenAI governance needs multiple layers beyond written policies, approved vendors, or a single security tool.

3.1 The Six Level Governance Framework · 3.2 Strategy & Policy (Level 1) · 3.2.1 Key artifacts · 3.2.2 MediAssist at Level 1 · 3.3 Risk & Impact Assessment (Level 2) · 3.3.1 Key artifacts · 3.3.2 Key Metrics · 3.3.3 MediAssist at Level 2 · 3.4 Design & Implementation Reviews (Level 3) · 3.4.1 Key Artifacts · 3.4.2 Key metrics · 3.4.3 MediAssist at Level 3 · 3.5 Launch Readiness Review (Level 4) · 3.5.1 Key artifacts · 3.5.2 Key Metrics · 3.5.3 MediAssist at Level 4 · 3.6 Operations & Monitoring (Level 5) · 3.6.1 Key artifacts · 3.6.2 Key Metrics · 3.6.3 MediAssist at Level 5 · 3.7 Learning & Improvement (Level 6) · 3.7.1 Key Artifacts · 3.7.2 Key Metrics · 3.8 Governing Agents at Runtime · 3.9 Summary

What you learn in chapter 3 →

Chapter 4: Securing GenAI

The chapter introduction explains why LLMs are vulnerable to instructions hidden in data, why filters cannot fully enforce the instruction–data boundary, and why greater AI capabilities raise the security stakes.

4.1 ClaimAssist · 4.2 Risk Factors · 4.2.1 ClaimAssist and How Risk Factors Stack · 4.3 Insecure Environments · 4.3.1 Insufficient Logging · 4.3.2 Vendor security gaps · 4.3.3 Supply-Chain Risks · 4.3.4 Uncontrolled usage and denial of service · 4.3.5 Improper secret management · 4.3.6 Shadow AI · 4.4 Model Risks · 4.4.1 Model Weight Risks · 4.4.2 Backdoors, Poisoning and Malicious Models · 4.4.3 Model Theft · 4.4.4 Mitigating Model-Level Risks · 4.5 Untrustworthy Input · 4.5.1 Beyond Chat-Based Attacks · 4.6 Data Security · 4.6.1 The Confused Deputy Problem · 4.6.2 When Existing Permissions Are Too Broad · 4.6.3 Poisoning the Well · 4.7 The Ability to Make Changes · 4.7.1 Building Controls for AI That Can Act · 4.8 Agency · 4.8.1 Why Autonomy Changes the Threat Landscape · 4.8.2 Agentic Browsers · 4.8.3 MCP and the New Tool Ecosystem · 4.8.4 Managing What Tools the Agent Can Use · 4.8.5 Tool Authorization · 4.8.6 When Humans Need to Stay in the Loop · 4.8.7 Containment and Detection · 4.8.8 Baseline and Advanced Practices · 4.9 ClaimAssist: From FAQ Bot to Agentic System · 4.10 Summary

What you learn in chapter 4 →

Chapter 5: Privacy

The chapter introduces how GenAI expands privacy risks, organizes them into four pillars, and emphasizes evidence of how personal data is handled.

5.1 The Four Pillars of GenAI Privacy · 5.2 Collection and Purpose · 5.2.1 No Valid Legal Basis · 5.2.2 Overcollection · 5.2.3 Vendor’s Inappropriate Use of Organizational Data · 5.2.4 Adoption Model Differences · 5.3 Storage and Memorization · 5.3.1 Memorization and Membership Inference · 5.3.2 Insufficient deletion of embeddings · 5.3.3 Insufficient Deletion of (Meta)data · 5.3.4 Adoption Model Differences · 5.4 Output Integrity · 5.4.1 Hallucinations and Defamations · 5.4.2 Overreliance and Automated Decision Making · 5.4.3 Adoption Model Differences · 5.5 User Rights & Governance · 5.5.1 Making models forget and correct · 5.5.2 Data Subject Access Requests (DSAR) · 5.5.3 Transparency · 5.5.4 Adoption Model Differences · 5.6 Agentic AI and Privacy · 5.6.1 Purpose Limitation (and Data Minimization) Under Strain · 5.6.2 Lawful Basis Challenges for Autonomous Processing · 5.6.3 Storage Limitation and Erasure Rights in an Agent Context · 5.6.4 Governance and Operational Challenges in the Long Term · 5.6.5 Summary of Privacy Risks and Mitigations in Agentic AI · 5.7 Summary

What you learn in chapter 5 →

Chapter 6: Ethical and Trustworthy GenAI

The introduction explains why trustworthy GenAI requires ethical safeguards beyond security and privacy, with controls scaled to potential harm and maintained throughout the product’s life.

6.1 Accuracy & Reliability · 6.1.1 Truthfulness & Hallucinations · 6.1.2 Robustness · 6.1.3 Consistency of Answers · 6.2 Bias and Fairness · 6.2.1 Real-world incidents · 6.2.2 What “fair” actually means · 6.2.3 A practical bias-testing loop for LLM-based systems · 6.2.4 Bias controls · 6.3 Overreliance · 6.3.1 Overreliance Controls · 6.4 Transparency and Explainability · 6.4.1 Transparency controls · 6.5 Safety and Toxicity · 6.5.1 Safety Controls · 6.5.2 Red Teaming in Practice · 6.6 IP and Copyrighted Material · 6.6.1 Provenance & license Risk · 6.6.2 During Use: Sharing IP and Legal Protections · 6.6.3 After Release: Outputs, Leakage, and Distilled Models · 6.7 Ethics in Agentic and Multi-Agent Deployments · 6.7.1 Agentic AI controls · 6.8 Summary

What you learn in chapter 6 →

Chapter 7: Regulation and Compliance

The introduction explains why teams need to clarify their EU AI Act obligations and how prior governance work can prepare them for discussions with Legal.

7.1 The Regulatory Landscape · 7.2 Classifying Your System · 7.2.1 Is It an AI System at All? · 7.2.2 Prohibited practices · 7.2.3 High Risk Classification · 7.2.4 General Purpose AI (GPAI) Models · 7.2.5 Transparency and AI Literacy Obligations · 7.3 Roles and Duties · 7.3.1 Provider, deployer, and other roles · 7.3.2 When a deployer becomes a provider · 7.3.3 What high-risk providers must do · 7.3.4 What high-risk deployers must do · 7.3.5 What your model provider owes you · 7.3.6 Finding yourself on the map · 7.4 What you already have and what’s still missing · 7.4.1 Are your existing frameworks enough? · 7.4.2 The evidence pack for providers · 7.4.3 The evidence pack for deployers · 7.5 Living Compliance · 7.5.1 Post-market monitoring · 7.5.2 Incident reporting · 7.5.3 Regulatory sandboxes and SME provisions · 7.6 ClaimAssist across the risk thresholds · 7.7 Summary

What you learn in chapter 7 →

Chapter 8: Governing for Uncertainty

The introduction uses cloud computing’s governance history to frame how organizations can govern GenAI despite unresolved risks and changing standards.

8.1 What can you govern today · 8.2 What remains unsolved · 8.3 Better tools are arriving · 8.4 The road ahead · 8.5 Summary

What you learn in chapter 8 →

Concepts explained in this book

These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.