An organization is scaling multiple AI initiatives across various departments. Data flows smoothly into the platform and passes initial validation…

Data lineage controls must be established at the point of data ingestion and initial validation. This is the first stage within the AI data architecture where metadata about the data's origin, ownership, and transformation history can be captured. Implementing traceability at this entry point ensures that a complete and auditable trail is maintained throughout the entire AI lifecycle, from raw data to model output. Waiting until data is curated or consumed by models means the crucial link to the original source is already lost, making it impossible to achieve end-to-end transparency.

A. This is too late in the lifecycle; transformations have already occurred, and the original data context is lost, making lineage tracking incomplete.

C. This stage is post-ingestion and transformation; establishing controls here would miss the critical initial steps, failing to capture the full lineage.

D. While data originates here, the control for the AI architecture must be established when data enters that specific environment, not within the source system itself.

1. Mendon, S., et al. (2021). MLOps: A Framework for Operationalization of Machine Learning Solutions. In: Proceedings of the 2021 International Conference on Information Technology (ICIT). Page 2, Section III-A, "The first step is to ingest data from various data sources... Data lineage and data versioning are important aspects to be considered." DOI: 10.1109/ICIT52682.2021.9672429.

2. Google Cloud. (2023). MLOps: Continuous delivery and automation pipelines in machine learning. "To ensure data quality, you should profile and validate your data... You should also track the lineage of your data to ensure accountability and reproducibility of your ML workflow." Section: Data processing.

3. Carnegie Mellon University, Software Engineering Institute. (2021). A CISO’s Guide to Artificial Intelligence and Machine Learning. CMU/SEI-2021-SR-011. Page 13, "Data provenance, also known as data lineage, is metadata that provides information about the origin and history of the data... Provenance should be captured at the time of data collection."