AI governance refers to the ensemble of laws, regulations, standards, voluntary commitments, and organizational processes that guide how AI systems are developed, deployed, monitored, and held accountable. It operates at multiple levels: international bodies set broad principles, national governments enact binding legislation (most notably the EU AI Act, which classifies systems by risk tier and mandates conformity assessments), standards organizations publish technical frameworks (NIST AI RMF, ISO/IEC 42001), and individual organizations implement internal review boards, model cards, and incident-response procedures.

Think of it like traffic law applied to a new class of vehicle. Just as the introduction of automobiles eventually required speed limits, licensing, insurance mandates, and road design standards, AI governance is society's attempt to define the rules of the road for increasingly capable automated systems.

A practical governance artefact is the model card—a short document published alongside a model release that records intended use cases, out-of-scope uses, training data sources, evaluation results, and known limitations. Regulators increasingly treat such documentation as a minimum baseline for compliance, and the EU AI Act requires conformity assessments and technical documentation for high-risk applications such as credit scoring or medical diagnosis.

AI governance draws on AI Ethics (which provides the normative foundation), AI Alignment (the technical effort to make systems comply with stated goals), AI Safety (risk identification and mitigation), Bias (a primary governance concern in high-stakes deployments), and Hallucination (a reliability risk that governance frameworks require disclosing).

You will encounter AI governance in regulatory filings, product compliance documentation, procurement requirements for public-sector AI, and in press coverage of legislative debates in the EU, US, UK, and China.