Part 7 — Governance, Compliance, and Safety

Overview

Governance is not a constraint imposed on agentic AI from outside. It is the engineering that makes agentic AI safe to operate at the autonomy levels organisations actually need.

Part 7 addresses the governance of deployed agent systems — not in the abstract, but as a set of concrete operational practices that must be designed, built, and sustained alongside the agents themselves. The chapters in this part cover what it means to hold agents accountable in practice: how you measure their behaviour once they are live, how you navigate the expanding body of law that governs their operation, and how you combine testing, oversight, and audit into an integrated governance model that can withstand regulatory scrutiny and operational stress.

The sequence reflects a deliberate logic. Governance begins with measurement — you cannot govern what you cannot see, and accountability without metrics is an intention rather than a practice. It then addresses the legal and regulatory context that defines the minimum obligations every deployment must meet. It ends with the operational architecture of governance itself: the testing methods, oversight mechanisms, and audit trail construction that translate good intentions into auditable, defensible practice.

Chapter 23 establishes the accountability foundation — the operational metrics, feedback loops, and continuous improvement cycles that keep deployed agents performing reliably as models update, usage patterns shift, and edge cases accumulate. Chapter 24 maps the regulatory landscape across jurisdictions and sectors, from the EU AI Act to sector-specific obligations in healthcare, financial services, and employment — the legal terrain that every deployment team needs to understand before going live. Chapter 25 brings testing, oversight, and audit trail construction together into a single operational governance model for live agent systems — the complete picture of what governing a running agent actually requires.

---

Chapters in This Part

Chapter	Title	Theme
23	Keeping Agents Accountable: Metrics, Feedback, and Improvement	Operational accountability
24	Navigating the Regulatory Maze: AI Compliance Across Jurisdictions	Legal and regulatory compliance
25	Governing Live Agents: Testing, Oversight, and the Audit Trail	Integrated governance architecture

---

Chapter 23 establishes the measurement foundation that Chapters 24 and 25 build on. Chapter 25 is the most operationally detailed chapter in this part and is particularly relevant for teams preparing for regulatory review or audit.