← AI Terminology
AI Governance
AI governance is the set of laws, regulations, standards, frameworks, and voluntary norms used to guide the development, deployment, and oversight of AI systems.
It operates at national (EU AI Act, US Executive Orders), international (G7, UN), and organisational levels.
It operates at national (EU AI Act, US Executive Orders), international (G7, UN), and organisational levels.
Why It Matters in AI
AI governance determines who bears legal responsibility when AI systems cause harm, which uses are prohibited, and what safety testing is required before deployment. For AI companies, governance shapes product design, compliance cost, and market access. For society, it is the mechanism through which democratic values — fairness, privacy, accountability — are embedded in technology that increasingly makes consequential decisions.
Key Points
| Aspect | Description |
|---|---|
| China | Mandatory algorithmic recommendation and generative AI regulations since 2022–23 |
| EU AI Act | World's first comprehensive AI law — risk-tiered (unacceptable/high/limited/minimal), in force 2024 |
| Voluntary | Model cards, datasheets, RSPs (responsible scaling policies) by labs |
| US approach | Executive Order on AI (Oct 2023), NIST AI Risk Management Framework — voluntary + sector rules |
| Key concepts | Risk tiering, conformity assessment, transparency obligations, human oversight requirements |
| Investment angle | Compliance costs are a moat for large AI companies; regulation can reshape competitive dynamics |
Simple Analogy
AI governance is to AI what aviation safety regulation is to aircraft: you don't ban flying, but you mandate airworthiness standards, pilot licensing, black-box recorders, and accident investigation — because the consequences of failure are too large to leave entirely to market incentives.
Common Usage Examples
- EU AI Act: real-time biometric surveillance in public spaces classified "unacceptable risk" and banned
- NIST AI RMF: companies map AI risks and document mitigations — widely adopted in US federal procurement
- OpenAI, Google DeepMind, Anthropic: all publish voluntary safety commitments (responsible scaling policies)
- G7 Hiroshima Process: international code of conduct for advanced AI developers
- Model cards: structured disclosure of training data, limitations, and intended use for each model release
Summary
In short: AI governance is the rule-making that determines what AI companies must do, cannot do, and must prove before their systems reach users — and it is moving fast.