Data governance strategy for AI
What is Data governance strategy for AI?
A plan that sets who owns data AI relies on, how its quality is checked and how every AI output can be traced back to its source. AI is only as reliable as data behind it. A data governance strategy for AI assigns an owner to each dataset and sets quality rules such as accuracy, completeness and currency, and records lineage from source to model to output. It also defines which data is approved for which AI use.
It extends existing data governance to cover training sets, retrieval sources and AI-generated data.
Why organisations need Data governance strategy for AI?
When an AI system gives a wrong answer, you must be able to understand where that answer came from, and who owned the process. Without lineage, nobody can tell. And without owners, nobody fixes it. Poor data can also produce outdated or biased outputs that reach customers.
APRA’s guide on managing data risk, CPG 235, treats data quality as something to manage across its life cycle. Its April 2026 AI letter outlines an expectation of assurance that covers data governance alongside cyber security and model performance.
How to build Data governance strategy for AI?
- List datasets each AI system uses.
- Assign an accountable owner to each.
- Define quality rules and automate checks.
- Record lineage from source to output.
- Approve datasets for specific AI uses, and review on every change.
