Data & Information ยท DAT-04
Finding and understanding data
Data available for AI is catalogued and described well enough that people can find it, understand what it means and its limits, and use it correctly.
Why it matters
Teams that cannot find suitable data either give up or use the wrong data. Descriptions of meaning, coverage and known limits stop data being used for purposes it does not support, which is a common cause of misleading AI outputs.
Levels
| 0 Absent | Confirmed absent: the evidence shows nothing is in place. Not known is recorded as insufficient evidence, not 0. |
|---|---|
| 1 Initial | Knowledge of what data exists sits with individuals; finding data means asking around. |
| 2 Developing | Some data sets are listed or documented, but descriptions are incomplete, out of date or held in separate places. |
| 3 Defined | A maintained catalogue describes data used or available for AI, including its owner, meaning, limits, conditions of use and sensitivity. |
| 4 Managed | Catalogue completeness and accuracy are measured, and use of the catalogue is tracked to show whether people find what they need. |
| 5 Optimising | Catalogue entries are kept current largely as a by-product of normal data work, and descriptions are improved from feedback by the people using them. |
Level 3 criteria
Level 3 is met when every applicable criterion is met. Criteria marked with an overlay apply only to organisations with that characteristic (overlays).
- DAT-04.1 A maintained catalogue covers data used or available for AI.
- DAT-04.2 Each catalogue entry records the owner, meaning, coverage, known limits and conditions of use.
- DAT-04.3 Each catalogue entry records its sensitivity classification and whether it holds personal data.
Assessment questions
- How would a team planning a new use of AI find out what data exists and whether they can use it?
- For a data set used by AI, where are its meaning, coverage and known limits written down?
- How do you know the catalogue is complete and current?
- The data catalogue, with sample entries for data used by AI
- The process for adding and updating catalogue entries
How to reach Level 3
- Choose a catalogue tool or a simple shared register, and give it an owner.
- Add every data set used or available for AI, recording owner, meaning, coverage, known limits and conditions of use.
- Record each entry's sensitivity classification and whether it holds personal data.
- Set a routine for owners or stewards (DAT-02) to keep entries current.
Common pitfalls
- Building a large catalogue that nobody maintains, so it goes out of date within months.
- Leaving out known limits, which are what stop data being misused.
Starter artefacts
- Data catalogue or register
- Catalogue entry template
Ownership and sequence
- Accountable
- data owner
- Contributors
- delivery lead, ai governance lead
- Depends on
- DAT-02
- Roadmap step
- 2
- Lifecycle
- design, acquire_build, operate
Cross-references
| AI Cyber Security Code of Practice | 5.1 partial |
|---|---|
| Algorithmic Transparency Recording Standard | Tier 2 Development data specification partial |
| ISO/IEC 42001 | A.4.3 partial |
| MITRE AI Maturity Model | P5.1 partial; P5.4 partial |
| UK Data and AI Ethics Framework | Transparency 6 partial; SAT 1.5 direct |
| UNESCO AI Maturity Framework | 6.1 direct; 6.5 partial |
direct: this control at Level 3 meets the item. partial: it covers part of it.