ISO 42001 Training Data Bias Checklist

How This Template Accelerates Your Audit & Career

A checklist for assessing a training or evaluation dataset for bias before a model learns from it. Seven areas: whether the served population is represented, whether the labels are fair and consistent, whether sensitive

attributes or their proxies are present, whether the data carries past decisions and access barriers, whether measurement and sampling differ between groups, whether the mitigation has been applied and measured, and sign-off. Run it once per dataset version, with the data steward and the model owner together. Mark each