Data stewardship
Taking responsible, ongoing care of data assets, keeping them accurate, well-described, and handled in line with the organisation's rules and values.
The Simple Version
Taking responsible, ongoing care of data assets, keeping them accurate, well-described, and handled in line with the organisation's rules and values.
Detailed Explanation
Data stewardship is the collective activity performed by data stewards, data owners, and governance committees to oversee data assets. It encompasses quality management, metadata maintenance, access governance, lifecycle management, and compliance monitoring. In the context of AI, stewardship extends to overseeing training and inference data throughout the model lifecycle, a dimension highlighted in ISO/IEC 5259 Part 5, which frames data quality governance as an ongoing stewardship responsibility rather than a one-time compliance task.
Key Characteristics
- Ongoing rather than episodic, stewardship is continuous across the data lifecycle
- Encompasses quality, metadata, access, and compliance dimensions
- Involves collaboration between business, technical, legal, and ethics stakeholders
- Produces a culture of data responsibility alongside operational controls
Why It Matters
Embedding stewardship into AI development teams reduces the risk of training on poor-quality, biased, or non-compliant data, and provides the audit trail that regulators and auditors look for.
Real-World Analogy
Like the stewardship of a national archive, not just storing records but ensuring they are accessible, authentic, well-described, and preserved for future use.
Common Misconceptions
- Stewardship is only about data quality, it equally covers metadata, access control, lineage documentation, and lifecycle decisions.
- Stewardship ends when data is collected, it continues throughout the full lifecycle including use, sharing, and eventual deletion.