# What is AI governance?

AI governance is how an organisation decides what its AI systems may do, who answers for them, and how that is checked. It turns laws and principles into owners, rules, controls and records.

## What it consists of

- **An inventory.** Which AI systems exist, what each does and what data it touches. Tools nobody approved count too: see [shadow AI](/glossary/shadow-ai).
- **Owners.** A named person for each system.
- **Rules.** What each system may and may not do, written down.
- **Risk assessment.** How likely and how bad a failure is, before launch and after changes.
- **Controls.** What enforces the rules while the system runs.
- **Records.** Evidence of what happened and what was done about it: see [audit logs](/glossary/audit-logs).
- **A plan for incidents.** See [AI incident response](/glossary/ai-incident-response).

## Governance and compliance

[Compliance](/glossary/ai-compliance) is meeting specific outside rules, such as the [EU AI Act](/glossary/eu-ai-act). Governance is the organisation's own system for deciding and checking, which is what makes compliance possible. Frameworks such as the [NIST AI RMF](/glossary/nist-ai-rmf) and [ISO/IEC 42001](/glossary/iso-42001) describe how to set it up.

## Where it usually fails

In the gap between the written rule and the running system. A policy document says the assistant never gives financial advice. Nothing in production checks. Governance that stops at documents is a statement of intent.

## With White Circle

[Control](/products/control) enforces written policies on every prompt, response and tool call, and keeps a record of what was flagged and why. [Analyze](/products/analyze) shows how those policies are doing over time.

