Data for AI Systems
Description
Focuses on data management for AI including quality assurance, provenance tracking, privacy protection, bias mitigation, and secure handling throughout the AI lifecycle.
Controls
| ID | Control | Description |
|---|---|---|
| A.6.1 | Data quality for AI systems | Processes ensuring AI training and operational data meet quality standards for accuracy, completeness, consistency, and ... |
| A.6.2 | Data provenance and traceability | Documentation of data sources, collection methods, transformations, and lineage throughout the AI system lifecycle. |
| A.6.3 | Privacy and personal data protection in AI | Privacy-preserving techniques and compliance with data protection regulations in AI data collection, processing, and sto... |
| A.6.4 | Data bias identification and mitigation | Systematic assessment and mitigation of bias in training data that could lead to discriminatory AI system outcomes. |
| A.6.5 | Data handling and security for AI | Secure data handling practices including access control, encryption, sanitization, and protection of training data and m... |
Implementation Guidance
Implement data quality frameworks with automated validation pipelines checking for completeness, outliers, and label accuracy. Maintain data lineage tracking from source through all transformations to model deployment. Apply privacy-enhancing technologies (differential privacy, federated learning) for sensitive data. Conduct fairness audits on training datasets across protected attributes with mitigation strategies for identified biases. Classify and protect AI datasets with encryption, access controls, and secure disposal procedures.
Evidence Requirements
Data quality assessment reports
Data lineage documentation
Privacy impact assessments
Bias audit results and mitigation plans
Data security controls documentation