Glossary

AI governance

AI governance is the system of authority, accountability, policies, processes, and oversight through which an organization directs and controls the development, deployment, and use of AI — including decision rights, documentation, escalation paths, and assurance activities. Governance sets who may approve what under which rules; it is related to, but not identical with, AI risk management, which focuses on identifying, assessing, treating, and monitoring specific risks.


Why it matters

Without governance, AI initiatives accumulate as pilots with unclear owners, inconsistent evaluation, and silent production use. Governance answers: who decides model and vendor choices, what evidence is required before launch, how exceptions are granted, and how incidents escalate — especially when agents can take actions, not only generate text.

Voluntary frameworks and standards (for example the NIST AI Risk Management Framework and ISO/IEC 42001 AI management-system guidance) inform governance design. They are not automatically legal obligations; organizations must map frameworks to their own risk appetite, sector rules, and contractual duties without treating “framework alignment” as proof of compliance.

Hypothetical illustration: An enterprise creates an AI governance board with authority to approve high-impact use cases, require model cards and evaluation packs, and revoke production access when controls lapse. Day-to-day risk assessments and mitigations still run under AI risk management procedures owned by product and security — governance sets the decision rights; risk management executes the risk cycle.

Mechanisms

  • Decision rights and accountability — clear owners for approval, operation, and incident response.
  • Policy and standards stack — acceptable-use, data, vendor, evaluation, and human-oversight requirements.
  • Assurance and oversight — reviews, audits, metrics, and escalation that make policy enforceable.

Boundaries

  • AI governance ≠ AI risk management as exact synonyms (governance = authority/oversight system; risk management = risk identification–treatment cycle).
  • AI governance ≠ “AI oversight” as an exact alias — oversight activities are part of governance, but the term is broader and often used loosely.
  • AI governance ≠ ethics slogans without enforceable decision rights and evidence requirements.
  • Aligning to a voluntary framework ≠ legal compliance by default.
  • Governance of AI tools ≠ the entire operating model, though it should connect to operating-model and agentic decision-rights design.

Sources

  1. National Institute of Standards and Technology (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0). NIST AI 100-1 — voluntary risk-management and governance-oriented guidance for AI systems.
  2. ISO/IEC 42001:2023 — Information technology — Artificial intelligence — Management system — requirements for an AI management system (standard; adoption is organizationally chosen unless mandated elsewhere).
  3. OECD — OECD AI Principles (2019; updates as published) — intergovernmental principles for trustworthy AI; not a private-sector compliance code by themselves.



Further research


Continue exploring

A short path from this concept into related research and practical resources.

  1. Related conceptAI risk management
  2. FrameworkAgentic Operating Model
  3. Deeper researchThe Agentic Operating Model Building
  4. Apply in DiscoverThe Enterprise AI Playbook Library
AI Governance — Definition for Enterprises | The Strategy Stack