Datasets: see what is under Personal data, and what the AI sees
Short answer
Under Personal data in the menu you find all your datasets, with area, number of records and whether the use has been recorded. Open a dataset to see the records with the real values, what the AI sees, the recognition risk (low, medium, high) for each record with suggestions for sending less, and the reports the AI has saved. Uploaded datasets can be deleted.
This is the real system with fictitious data. Try the walkthroughs in the articles to click through.
A dataset
This is the real system with fictitious data. Try the walkthroughs in the articles to click through.
Recognition risk
Even without a name, a person can be recognisable if they are the only one with a particular combination, e.g. age group, region and job title. For tables about people, GoalCube therefore gives each record a level:
| Level | Means |
|---|---|
| Low | At least four other records have the same combination. |
| Medium | The record shares the combination with 1–3 others. |
| High | The record is the only one with its combination. The AI may be able to work out who it is. |
The box Can anyone be recognised from the columns? on the dataset's page shows how many records stand out, and on which columns. If there are records with a high risk, GoalCube suggests what you can send less of, e.g. a range instead of length of service, a region instead of a city, or not sending a column at all. An uploaded dataset cannot be edited afterwards: delete it and upload the file again with the suggestions. "No records stand out" means a lower risk – not no risk.
The AI only gets the level for each record (low, medium or high), never which columns make the record unique.
Words that can point to a person
If you made a choice for the yellow words at upload, the dataset's page shows how many words were found, and how many were made broader or hidden. The choices are made at upload; see Upload a table.
Delete a dataset
Uploaded datasets can be deleted from the dataset's page. The deletion is logged. Example datasets cannot be deleted.
Screens recorded from GoalCube 0.26.0 on 22 October 2026 with fictitious data.