AI glossary · Basic · Data engineering · Safety & alignment
Data governance
Data governance is a framework that defines, and enforces, who manages and uses an organization's data, under what rules and how. It covers data owners, access rights, quality standards, retention and deletion rules, and more.
About this term
- Level
- Basic · Technical terms you often meet in AI news
- Category
- Data, ops & platforms
- Added
- 2026-10-08
How it works
First, an owner (data owner) is assigned to each dataset, and a definition of what that data means is documented. Next, access rights are divided according to who can see which data, and sensitive data such as personal information is marked separately and handled more strictly. Where data comes from and where it flows (data lineage) is tracked, and quality standards are set and checked regularly. These rules are set by a committee or a dedicated team and applied to real systems with tools such as a data catalog.
Basic governance checklist
Ownership and definitions
- Does each dataset have an assigned owner?
- Is the meaning of the data documented?
Protection and access
- Is it clear who can see which data?
- Is sensitive data marked separately?
Tracking and quality
- Is it tracked where data comes from and where it goes?
- Are quality standards checked regularly?
Why it matters
An AI model is only as good as the data it learns from and refers to. If a model is built on data of unclear origin or data used without permission, its results are hard to trust, and this can lead to personal data leaks or copyright problems. Also, if departments define the same metric differently, the same question gets different answers. As AI and privacy regulations tighten in many countries, a framework that can explain what data was used and how is becoming a prerequisite for doing business.
What to keep in mind
Governance is not finished by adopting a single tool; people, processes and technology have to work together. If the rules are too strict, people who want to use data get worn out by the procedures and look for workarounds. That is why it is often applied in stages, starting with the highest-risk data. Activities such as data quality management and de-identification of personal data can be seen as detailed activities within the broader framework of governance.
Example
When a company builds a RAG chatbot that answers questions based on internal documents, without data governance even documents meant for only some employees, such as HR records or executive reports, could be exposed to all users as sources for answers. So many companies attach a security level and access rights to each document and design the retrieval step to pull only documents that match the permissions of the person asking.