Data governance is the coordinated use of policies, processes, roles, and controls to determine how data is collected, managed, used, protected, and overseen. In AI, it extends beyond individual consent to systematic oversight of who controls data, how it moves through the lifecycle, and whether safeguards remain adequate. section 3.18
AI creates governance questions at each stage of the data lifecycle:
- Collection can involve public information gathered without clear consent or provenance. section 3.18
- Curation can introduce or amplify bias when datasets are undocumented, unbalanced, or insufficiently reviewed. section 3.18
- Training can transform raw data into representations that are difficult to trace or delete. section 3.18
- Outputs can expose private details or reproduce biased assumptions embedded in training data. section 3.18
Organizations respond with data inventories, documentation standards such as datasheets and model cards, red-teaming protocols, bias assessments, and AI-specific privacy impact assessments. These mechanisms make accountability more systematic, but scale, opacity, and velocity still make AI data difficult to govern effectively. section 3.18
Education and generative AI
Schools must consider whether uploading student assignments or personal details to an AI system exposes protected information to an unauthorized provider. Practical responses include reviewing vendor agreements, establishing consent and data-governance protocols, updating acceptable-use policies, and training teachers. Existing privacy laws provide important protections, but adaptive and generative systems create gaps that require transparency and accountability. section 3.16
