Article 10 requires that training, validation, and testing datasets for high-risk AI systems meet quality criteria. The Data Governance module tracks your datasets and their governance measures.
Adding a dataset
Go to AI Act → Data Governance and click Add Dataset.
| Field | Required | Description |
|---|---|---|
| Dataset Name | Required | Identifier for the dataset |
| AI System | Required | Which AI system uses this dataset |
| Type | Required | Training, Validation, or Testing |
| Description | Optional | What the dataset contains and how it was collected |
| Size | Optional | Number of records or volume |
| Quality Measures | Optional | What quality controls are in place (bias testing, completeness checks, etc.) |
Art. 10 specifically requires datasets to be relevant, representative, free of errors, and complete. Document your bias detection and mitigation processes here.